Radiochim. Acta 95, 477–492 (2007) / DOI 10.1524/ract.2007.95.8.

477
© by Oldenbourg Wissenschaftsverlag, München

Optimization of the decision threshold for single radioactive

counting

By A. Vivier1 and J. Aupiais2 , ∗

1
CEA, Institut National des Sciences Techniques et Nucléaires, Centre de Saclay, 91191 Gif-sur-Yvette, France
2
CEA, Département Analyse et Surveillance de l’Environnement, Centre de Bruyères-le-Châtel, BP 12, 91680 Bruyères-le-Châtel, France

(Received September 19, 2006; accepted in revised form February 22, 2007)

Decision threshold / Bayesian statistics / cision threshold and detection limit have been elaborated
Low radioactive count rate / Nuclear counting based on laws of distribution, generally the Gauss or the
Poisson distribution. The detection limit is calculated from
the decision threshold and assuming that only statistical fluc-
Summary. When the activity of a sample is close to the tuations (type A uncertainty) are the principal sources of
background level, the decision threshold is classically defined fluctuation [1]. Recently, a new approach has been used
by considering distributions of sample and background as which takes into account the type B uncertainties (due to
being equal. Recently, the Bayesian approach has been con- the lack of knowledge of the chemical yield, the sampling,
sidered in the standard ISO to refine the determination of
the efficiency of detection, etc.) [2]. Based on Bayes’ theo-
the decision threshold by taking into account all accessible
information prior to measurement such as type A and type rem, this new approach has led to the redaction of new
B uncertainties. However, simplifications using Gaussian ap- international standards for the measurement of radioactiv-
proximation and experimental values instead of true means ity, namely ISO 11929 [3–6]. Nevertheless, this standard
are often used to facilitate calculations. In this paper, we brings new constraints for laboratories such as i) the control
develop a complete treatment without simplification, based of all sources of uncertainties, ii) a well-known background
on the Bayesian approach and Poisson distribution. Minimal and iii) at least one event in the region of interest during
informations have been considered: one single raw counting the sample measurement. In this paper, we propose to build
for the sample and one previously acquired background. From a decision threshold by taking into account all informations
one single background counting, the net background probabil- actually available prior to sample counting, namely one sin-
ity law is calculated and then a decision threshold is deduced.
gle background counting. Indeed, due to the long time dedi-
In particular, we demonstrate that the decision threshold is
defined for any case including very low background or even cated to environmental measurements, the determination of
null event. Comparisons with classical approach as well as the background can not be performed before each sample
the Bayesian treatment in the new ISO 11929 have been measurement. The single determination of the background
carried out. Applications of this decision threshold for the is generally justified by its temporal invariance (except may
optimisation of radioactive measurement or in case of a set be for the γ -spectrometry where the background depends
of minimal detectable activities used to determine average mainly on the cosmic rays and/or the sample itself [7]).
releases are given. Bayesian treatment also gives relevant Basically, for nuclear counting the notion of the deci-
informations such as the probability for a source to be ra- sion threshold appears when the value of the raw counting
dioactive when the net number of counts is below the decision S of a sample suspected of containing radioactive matter is
threshold.
similar to the value B of a preliminary counting of the back-
ground without the presence of the sample. The condition
1. Introduction for the net counting S − B > 0 is not sufficient to decide
whether a sample is radioactive or not. It is then necessary
Radioactive measurement in environmental samples needs to use statistical hypotheses under which a decision thresh-
very sensitive techniques because the concentration of ra- old can be built and the following question will be solved:
dioisotopes may be below the picogram per gram of matrix. Can you consider that the net counting value of the sample
For short-lived radioisotopes nuclear spectrometric tech- is significantly important so that the presence of radioactive
niques such as γ -, α-spectrometry and liquid scintillation matter is certain? The decision threshold is associated with
counting are often used. When the measurement tends to the the probability of an error of the first kind, or error α, that
limit of sensitivity, the number of events collected is of the a positive answer to the previous question after using a sta-
same order of magnitude as the background. Yet, the ana- tistical test may be wrong. The construction of a statistical
lyst has to answer whether a radioactive element is present test according to the Neyman-Pearson method is based on
or not. In order to answer this question, the notions of de- the following process: 1) determination of a discriminative
statistics D, 2) choice of a preferred hypothesis H0 associ-
*Author for correspondence (E-Mail: [email protected]). ated with an average value D of D. This hypothesis called
478 A. Vivier and J. Aupiais

the null hypothesis is generally chosen such as D = 0, 3) for the background between its acquisition and that of the
construction of the distribution of all possible experimen- sample. Most of the time, the analyst assesses the true mean
tal values of D due to experimental uncertainties according by the acquisition of one background. This single determin-
to the hypothesis H0 , 4) determination of the interval of re- ation is generally considered as being equal to the true mean,
ject, which defines the critical range, as a function of the which is naturally not the case. This approach can even come
error α chosen by the operator; and the decision thresh- to a deadlock if no count is observed; some spectrometers
old, 5) comparison between the experimental determination may reach this sensitivity in a few cases. Nevertheless, the
Dexp and the decision threshold. This comparison leads to probabilistic evaluation of the true mean is possible by ap-
the reject of the preferred hypothesis if Dexp is within the plying the Bayes’ theorem. For the purpose of clarity, we
critical range (this decision is associated with the error α) will first consider same counting times ts and t0 , i.e. n = 1.
or its acceptance if outside the critical range (this decision The results will be then generalized.
should be associated with the error of the second kind, or
error β, itself associated with the choice of an alternative
hypothesis H1 ). When considering nuclear counting, the dis-
criminative statistics is usually the net counting S − B and 2.1.1 Distribution of the probability of the true mean
the null hypothesis H0 is “the sample is not radioactive”. The in nuclear counting knowing only a single measurement
distribution of D according to H0 is built as the difference P(µ|N1 )
of two variables having the same distribution, namely that of
the experimental background. This construction necessary The distribution of the nuclear disintegrations and the back-
involves the experimental uncertainties associated with the ground follows a distribution of Poisson which is summa-
determination of the true mean of the background knowing rized as “the probability to obtain N counts knowing that the
that it is generally determined by only one measurement. In true mean is µ”:
order to include the experimental uncertainty, it is then ne- e−µµN
cessary to apply the Bayesian approach. P(N|µ) = . (2)
N!
N plays the role of a random parameter and can take all
2. Theoretical part integer values between 0 and +∞. It should be noted that
Let consider a random variable N associated with a law N is variable and µ constant. Practically, when consider-
of probability P(N = N0 ). We will decompose in this pa- ing only one single counting, one particular value of N is
per this variable into N = µ + ε, where µ is the true mean known while µ is unknown. It is the reverse problem where
of N and ε a centred residual random variable. As ex- N is constant and µ variable. Therefore, the true mean µ can
ample, the Poisson distribution is usually applied for nu- be considered probabilistically as a random variable. The
clear counting. Thus, according to the previous description application of the Bayes formula allows determining the dis-
of N, we get E(N) = µ and V(N) = V(ε) = σ 2 = µ. The tribution of µ knowing the value of one unique counting N1 :
evaluation of the uncertainty of the standard deviation for
a Poisson distribution obtained thanks to a single counting e−µ µN1
P(µ|N1 ) = . (3)
is always equal

to0.5. Indeed,

the uncertainty is expressed N1 !
2 2
dσ 1
by V(dσ) = σ =
2
√ N = 0.52 when consid- Although both right sides of Eq. (3) and Eq. (2) are identi-
dN 2 N
ering V(dσ) as the variance of the residual fluctuations dN cal, Eq. (3) is not a Poisson law since the roles of random
of N around µ. This property will be used further. and constant parameters are reversed. Also N1 may take only
integer value while µ is continuously variable. This law is
a density of probability called “γ function” [8] of parame-
2.1 Development of the hypothesis H0 ter N1 + 1 [9]. Indeed, Bayes’ theorem applied to the Poisson
Let define the hypothesis H0 : “the sample is not radioactive”. distribution consists in calculating the following expression:
The discriminative statistics D is the quantity:
P(N1 |µ)f(µ)
P(µ|N1 ) =  . (4)
Nt P(N1 |µ)f(µ)dµ
Nnet = Nts − 0 , (1) µ
n
with Nts as the raw counting of a non radioactive sample, The prior f(µ), considered as the density of probability of µ
Nt0 as the counting of background, ts as the counting time before any counting, is chosen as a constant to obey the max-
of the sample, t0 as the counting time associated with the imization of the information (or the entropy) [2, 10, 11]. The
determination of the background, and n the ratio t0 /ts . The choice of a flat prior may be subject to intense discussion,
law of distribution of Nts will be determined thanks to one but we argue here that without any complementary source of
information: the number of counts of the background N1 . information about the nature of the background before the
All variables in bold and italic are random while all vari- single measurement, a flat prior is justified.
ables in normal style are numbers experimentally obtained Eq. (4) is easily resolved by performing successive inte-
after a measurement. Minimal hypotheses are: Poisson dis- grations by parts. The obtained distribution is characterized
tribution of nuclear counting and background distribution, by: a) an expected value E(µ|N1 ) = N1 + 1, b) a variance
neglecting pile-up and assuming invariance of the true mean V(µ|N1 ) = N1 + 1. Indeed, the expected value is calculated
Optimization of the decision threshold for single radioactive counting 479

according to the following relations: probability to get N2 counts is built as the sum over all pos-
sible values of the true mean weighted by the probability that
∞ µ is the true mean knowing that N1 counting was observed:
e−µµN1
E(µ|N1 ) = µ dµ 
N1 !
0 P(N2 |N1 ) = P(N2 |µ) f(µ|N1 )dµ (8)
∞
1 µ
= e−µµN1 +1 dµ = N1 + 1 . (5) ∞
N1 ! e−µµN2 e−µµN1
0
P(N2 |N1 ) = × dµ
N2 ! N1 !
In the same way, the variance is equal to: 0
∞
1
V(µ|N1 ) = E(µ |N1 ) − E(µ|N1 ) .
2 2
(6) = e−2µµN1 +N2 dµ. (9)
N2 !N1 !
V(µ|N1 ) = (N1 + 2)(N1 + 1) − (N1 + 1)2 0

= (N1 + 1) . (7)
It yields:
The obtained variance is larger (V = N1 + 1) [9, 10, 12] than 1 (N1 + N2 )!
that of the Poisson distribution for the law of µ. It can be no- P(N2 |N1 ) = . (10)
2N1 +N2 +1 N2 !N1 !
ticed that for a particular condition where no event has been
detected (N1 = 0) the mean and the variance are equal to 1. By applying the theorem of the total mean (or total expecting
Therefore, it is not necessary, as suggested by Weise [11, 12] value), one can write:
to perform counting for sufficient time to have at least a few
counts. Thus, we can handle the true mean as a random E(N2 |N1 ) = E[E(N2 |µ|N1 )] = N1 + 1 . (11)
variable such as µ = (N1 + 1) + ε with E(µ) = N1 + 1 and
V(ε) = N1 + 1. It can be understood as an epistemic uncer- Similarly, the theorem of the total variance allows writing:
tainty related to the lack of knowledge of the true mean and
not to a real fluctuation. By definition, the epistemic uncer- V(N2 |N1 ) = E[V(N2 |µ|N1 )] + V[E(N2 |µ|N1 )] . (12)
tainty decreases in function of the amount of information of
the background. Since V(N2 |µ) = µ, then E[V(N2 |µ|N1 )] = E(µ|N1 ) = N1 +
1. Moreover E(N2 |µ) = µ. It comes V[E(N2 |µ|N1 )] =
2.1.2 Distribution of the probability in nuclear counting V(µ|N1 ) = N1 + 1.
of a forthcoming second counting knowing only a single Therefore,
measurement P(N2 |µ|N1 )
V(N2 |N1 ) = 2(N1 + 1) . (13)
Now, let N2 be a possible value of the number of counts asso-
ciated with the random variable N2 (= Nts ). The aim is here According to the previous convention, we write N2 = µ +
to determine the distribution of the probability of a future εstoc = (N1 + 1) + εepist + εstoc , with V(εepist ) = V[E(N2 |µ)]
counting knowing only one previous determination N1 . The as the epistemic uncertainty related to the true mean and

Fig. 1. Comparison between Poison

distribution for µ = 2 and µ = 50
(black distribution) and P(N2 /N1 ) for
N1 = 2 and N2 = 50 (white distribu-
tion).
480 A. Vivier and J. Aupiais

Fig. 2. Distributions of Nnet |N1 for

N1 = 10 and N1 = 2 (histograms)
and the corresponding centred
and reduced gaussian distribution
(straight line
√ & cross) of the same
variance 2(N1 + 1).

V(εstoc ) = E[V(N2 |µ)] as the stochastic variation of a nu- is necessary to calculate the standard deviation σH0 of the
clear counting around the true mean. Examples of distribu- distribution H0 .
tion are given in Fig. 1. Quantile values of the law N2 |N1 are Let decompose the random variables Nt0 and Nts as Nt0 =
given for several N1 values in the annexe (Table A). N1 + 1 + εepist,t0 + εstoc,t0 and Nts = N1 + 1 + εepist,ts + εstoc,ts . We
can write:
2.1.3 Distribution of the net background counting and
determination of the decision threshold Nnet = (εepist,t0 − εepist,ts ) + (εstoc,t0 − εstoc,ts ) (18)

In Eq. (1), both random variables Nt0 and Nts are defined The epistemic uncertainties are totally correlated because
using the Bayesian approach and are deduced from the sin- the true mean is the same for both variables Nt0 and Nts .
gle value N1 . The Bayes’ theorem allows writing: Since cov(εepist,t0 , εepist,ts ) = V(εepist,t0 ) = V(εepist,ts ), you can
deduced that:
∞
e−µ µ Nts e−µ µ Nt0 e−µ µN1
P(Nts , Nt0 |N1 ) = dµ . (14) V(εepist,t0 − εepist,ts ) =
Nts ! Nt0 ! N1 !
0 V(εepist,t0 ) + V(εepist,ts ) − 2cov(εepist,t0 , εepist,ts ) = 0 (19)
It yields:
The total variance is V(Nnet ) = 2(N1 + 1) despite the fact
1 (Nts + Nt0 + N1 )! that Nt0 and Nts have the same variance, the variance of their
P(Nts , Nt0 |N1 ) = . (15) difference is still equal to 2(N1 + 1) because of the covari-
3 Nts +Nt0 +N1 +1 Nts !Nt0 !N1 !
ance term. Although Eq. (16) is cumbersome to calculate,
The determination of the law of probability Nnet (which is in the convolution product can be numerically determined and
fact the distribution of the H0 hypothesis) is a discrete convo- the H0 distribution is obtained without approximation. The
lution product obtained by summing all probabilities of Nt0 determination of the decision threshold is then exact for any
and Nts for each determination of Nnet . N1 although we have observed a good correlation with the
Gauss distribution of null mean and standard deviation σ =
√
 1 (2N + Nnet + N1 )! 2(N + 1) (see Fig. 2). Thus the decision threshold is calcu-
P(Nnet |N1 ) = .
N≥Nnet
32N+ Nnet +N1 +1 N!(2N + Nnet )!N1 ! lated with
√ an excellent approximation applying Eq. (17) with
(16) σH0 = 2(N + 1). The Bayesian approach is therefore essen-
tial for very low background. Above N1 = 10 counts, differ-
A decision threshold γ can be deduced from Eq. (16) such as ence between the classical approach and the Bayesian one
P(Nnet ≥ γ) = α % and leads to the following equation: is below 5%, largely negligible compared to the uncertainty
associated with the decision threshold (14% for N1 = 10).
γ = k1−α σH0 . (17) In this case, the Bayesian approach has no particular in-
terest. Since the decision threshold is a random variable,
In the classical approach, a Gaussian distribution is consid- one can deduce by derivatization its associated uncertainty
ered allowing chosing k1−α from the standardized normal
distribution. In order to verify this useful approximation, it 1
A free and simple Excel worksheet is available upon request.
Optimization of the decision threshold for single radioactive counting 481

√ σγ 0.5
σγ = 0.5k1−α 2 (or its relative uncertainty =√ ).
γ N1 + 1
It must be noticed, even for N1 small (N1  = 0), that the de-
cision threshold based on either the classical or the Bayesian
approaches are not significant compared to its associated un- Fig. 3. Region  1 – Nnet < γ , activity declared not significant at β %
certainty. The classical approach, neglecting the epistemic to fail, region 
2 – no decision, continue the counting or perform an-
uncertainty, wrongly considers the distributions of Nt0 and other counting, region 3 – activity declared significant at α % to fail
and expressed with its uncertainty uNnet = Nnet + 2Nts .
Nts as
√independent variables of mean N1 and standard devia-
tion N1 . Thus, the variable Nnet = Nts −
√ Nt0 is characterised
by a mean 0 and a standard deviation 2N1 . The Bayesian the net counting is above the decision threshold Nnet ≥ γ ,
approach is therefore essential for the very low background the measurement is declared significant with a risk of error
since more coherent and precise. Indeed, when consider- equal to α and has
ing the raw counting Nraw = N1 + Nts , the decision threshold √ to be expressed with its associated un-
certainty unet = Nnet + 2N1 (ts ) (region ).
3 When the Nnet
based on the Bayesian approach is coherent with the one value ranges within the uncertainty associated with the de-
based on the net counting. For the classical approach this cision threshold, it is then recommended to perform another
coherence fails. measurement whenever possible or to continue the counting
(region ).
2
2.1.4 Increase of the amount of information about
the background
3.2 Generalisation with t0  = ts
If we assume the availability of additional information about
the background, e.g. n countings (N1 , N2 , ..., Nn ) during Let define the following hypotheses: the likelihood function
tS , the law of probability of n × µ is built as the likelihood of µ|N1 , and the Poisson distributions of Nt0 |µ and Nts |µ, the
function: application of Bayes’ theorem to the Eq. (1) with t0 = nts ,
gives:

n

e µ
n −µ Ni ( e−µ)n µNi ∞ µ

P(µ|N1 , N2 , ..., Nn ) = =
i=1 e− n µNts e−nµµNt0 e−nµ(nµ)N1
Ni ! 
n P(Nts , Nt0 |N1 ) = × × dµ
i=1 Ni ! Nts ! Nt0 ! N1 !
0
i=1

n (22)
Ni
e−nµ
µi=1
= 
n . (20) It yields:
Ni !
1 1
i=1
P(Nts , Nt0 |N1 ) =
(2 + ) 1 Nts +Nt0 +N1 +1
n Nts
The likelihood function is similar to the law associated with n

n
(Nts + Nt0 + N1 )!)
a single counting N1 (t0 ) = Ni during t0 (t0 = ntS ). One can × . (23)
i=1 Nts !Nt0 !N1
deduce:

 By considering the Eq. (1), the determination of the law of
1 1 1 probability Nnet |N1 is now:
V(µ) = V(nµ) = Nts + . (21)
n2 n n
 1
Eq. (21) expresses the decrease of the epistemic uncertainty P(Nnet |N1 ) =
( )
1
1 Nnet +Nt0 1+ n +N1 +1
with the increase of the amount of information during either N1 with
2+ n
n countings ts or a single counting with t0 = nts . N
Nnet + nt0 integer ≥ 0
1 Nnet + Nt0 1 + n1 + N1 !
× Nt × (24)
3. Applications – optimization of the decision n Nnet + n
0
Nnet +
Nt0
!Nt0 !N1 !
n
threshold
3.1 Use of the decision threshold, the detection limit The evaluation of the standard deviation associated with
and the corresponding uncertainties Eq. (24) needs to decompose the random variables Nt0 and
Nts into Nt0 = (N1 + 1) + εepist,t0 + εstoc,t0 and Nts = (N1n+1) +
The introduction of an uncertainty associated with the deci- εepist,ts + εstoc,ts . The epistemic variables are interdependent
sion threshold leads to define how such a criterion should be with a covariance term equal to n12 (N1 + 1). Finally, you can
used under actual operating conditions. If the determination demonstrate that the variance of Nnet is:
of the net counting gives a value below the decision thresh-
old (Nnet < γ ), then the measurement is considered as being 1 1
V(Nnet ) = (N1 + 1) + 2 (N1 + 1)
not significant with an error equal to the β error (region , 1 n


n 
Fig. 3). When accepting the H0 hypothesis, i.e. “the sample 1 1
= 1+ Nts + , (25)
is not radioactive ...”, this decision should be extended to n n
following assumption “... but if it is really radioactive, there
is β% of chance that its activity be at the most equal to an with Nts the background counting during ts evaluated by
activity which is in average equal to the detection limit”. If Nts = Nn1 . As a consequence, the decision threshold for the
482 A. Vivier and J. Aupiais

Gaussian approximation of Eq. (17) and the justifications quantiles is given. This point is of minor importance for
developed after, is: N high but is critical for N < 10 counts. According to data


 given in the annexe, it is now possible to assess such devia-
1 1 tion and to choose the adequate quantile. As N1 is a random
γ = k1−α 1+ Nts + (26) variable, the decision threshold is also a random variable.
n n
For n  = 1, the uncertainty associated with γ is:
In practise, the decision threshold
√ can be enhanced by in- 

creasing n (the optimal gain is 2). Nevertheless, above n = 1 1 1
uγ = k1−α 1+ . (27)
10, the improvement is no longer important. You √can notice 2 n n
that for n = 1, we find the expression γ = k1−α 2(N1 √ + 1).
For n  1, the decision threshold tends to γ = k1−α N1 . The relative standard deviation is
Due to the similarity of both distributions Nnet |N1 and Gauss,    
uγ 1 1 1 uγ
you can either chose the quantile of the Gauss distribution =√ ×  ≈√ × .
γ n=1 n 2 N +1 n γ n=1
k1−α or that of given in the Annexe – Table B k1−α (calcu- ts n
lated for n = 1). The data are restricted to N0 ≤ 30 because
the factorial terms above 70 give small probabilities which Thus, for instance, a ratio n = 10 decreases the decision
are rounded down to 0 and the sum of their contributions is threshold by 25% while its uncertainty is divided by a fac-
no longer negligible for N1 > 30. It can be noticed that rigor- tor 3. A few examples are shown in Table 2.
ously due to the discrete distribution of Nnet |N1 , the quantile
in the Annexe – Table B should be preferred. 3.3 Optimization of the minimal detectable activity
We have compared in Table 1, several equations of the The activity, expressed in Bq, is given by the relation
decision threshold based either on the conventional statistics Nnet
or the Bayesian approach. As expected, all equations tends A= , where η is the efficiency of detection and ts the
ηts
to the same form (Gauss distribution) when N is high. For counting time. When applied to a net counting equal to the
N small, differences are noticeable between the conventional decision threshold γ (expressed in s−1 ) one obtains a deci-
γ
approach and the Bayesian treatment. Nevertheless, these sion threshold γ ∗ expressed in Bq: γ ∗ = .
differences are not statistically significant since the cor- ηts
responding uncertainties associated to the decision thresh- The decision threshold can be
expressed  in function of
1 1
olds overlap. As example for N1 = 1, we obtain γ = 5.3 ± the background count rate rs = Nts + :
1.4 (conventional) and γ = 4.0 ± 1.4 (Bayesian). The other ts n
equations based on the Bayesian approach give the same 

equation when ts = t0 . Difference appears only for ts  = t0 due 1
γ =k 1+ r s ts , (28)
to a different approach. In particular, the approach developed n
by Weise and partially applied in the ISO norm 11929 does
and
not consider the variance of the net distribution according to 
the H0 hypothesis as being always equal to 2(N + 1). More- k 1 + n1 rs
over, the quantile of the Gauss distribution is always used γ∗ = √ , (29)
instead of an exact treatment and no evaluation between both η ts

Table 1. Comparison between several equations of the decision threshold involving conventional statistics and
Bayesian approach.

Origin t0  = ts t = t0 = ts t = t0 = ts and N large

Conventional, ⎛  ⎞   

2

Poisson k21−α
⎝1 + 1 + 2t 0 N 1 1 ⎠
k21−α 8N k1−α √
distribution + 1+ 1+ 2 2N
2ts k1−α ts t0 ts 2t k1−α t
DIN 25482-1, etc.


Bayesian,
N+1 1 1 k1−α k1−α √
Weise [12] k1−α + 2(N + 1) 2N
ts ts t0 t t



Bayesian,
k1−α ts ts k k1−α √
This work 1+ N+ 2(N + 1) 2N
ts t0 t0 t t
k = k1−α or k1−α


(Gaussian approximation)

Bayesian, 

ISO 11929 N 1 1 k1−α √ k1−α √
k1−α + 2N or 2N
ts ts t0 t t
k1−α
2(N + 1) if Nnet = 0
t
Optimization of the decision threshold for single radioactive counting 483

Table 2. Numerical examples and comparisons for a few values of the background during a time tS .
The quantiles used in Eq. (17) (see Annexe – Table B) and Eq. (26) correspond to the same error α.

n N1 N1 (ts ) α (%) γ Eq. (17) γ Eq. (26) uγ /γ (%)

measured estimated

1 0 0 4.03 3 2.5 37
1 1 1 4.11 4 3.5 28
1 5 5 5.44 6 5.6 18
1 10 10 5.39 8 7.5 14
5 0 0 6.69 1 0.7 37
5 5 1 5.45 2.2 1.9 18
5 25 5 5.07 4.4 4.1 9
5 50 10 5.19 6 5.7 6

The quantile can be either k1−α (satisfying approximation – the standard of reject. Two cases are considered: scenario
Gauss distribution) or k1−α (exact value – Nnet |N1 distribu- 1 – the addition of n measurements (net counting) carried
tion). Thus, from a practical point of view, when needing out identically on the same object (eventually except time
to lower the minimal detectable activity, several possibilities of counting), scenario 2 – measurements carried out on
can be considered: to decrease rs , to increase ts , n, ε or the α the same object but under different conditions (overall effi-
error (decrease of k). ciency, sampling).

3.4 Optimization of the acquisition time

3.5.1 Addition of n measurements under conditions
The management of nuclear measurements requires to con- of repeatability
sider the time allocated for the background acquisition (t0 )
For n measurements of a sample during various individ-
and that of allocated for j sample determinations ( j × ts )
ual times ti and all equal things in addition, in particu-
within a period of time t not extendable. On the basis of the
lar the efficiency of detection, one can individually deter-
relations t0 + jts = t and t0 = nts , the decision threshold is
mine a√net counting Ni = Bi − Si and a decision threshold
equal to:
γi = k 2Si (k = k1−α or k1−α ). These countings are equiva-
 lent to one single countingof a corresponding duration
k 1 + n1 rs 
t = ti , S = Bi and B = Si . A cumulated net counting
γ∗ =  . (30) i i  i
η n+1 j t N = Bi − Si = Ni , and a cumulated decision thresh-
i i  i 
Basically, the optimization consists in finding the value of old γ = k1−α 2 Si = γi2 are deduced. The decision
dγ i i
n which cancels the derivative . One obtains after calcu- threshold related to the addition of several measurements
dn √
√ t j is the quadratic sum of the partial thresholds of decision.
lations n = j and, finally: ts = √ and t0 = √ t. In some case, the addition of results enables to detect a ra-
j+ j j+ j
Some examples are given in Table 3. dioactive sample although each measurement gives negative
decision (Table 4).

3.5 Addition of results

The environmental monitoring within the regulatory frame- 3.5.2 Addition of several activities
work of nuclear facilities leads to determine the cumula- When the values of countings from the same object are not
tive total of all released activities and to compare them to cumulable (various efficiencies or different nuclear spectro-
metric techniques, etc.), the addition must be carried out
with results expressed in activity. Two cases must be distin-
Table 3. Optimization of the counting time, expressed as function of guished (Fig. 4).
the number of samples k. Case A: An unspecified partition is carried out and in-
dependent measurements for each part are performed. The
k ts / (ts + t0 ) t0 / (ts + t0 ) n = t0 /ts total activity of the initial object is the sum over its partial
activities. Let us suppose as example a sample shared into
1 0.5 0.5 1.0 two distinct parts. The overall activity, with t0 = tS , is simply
2 0.293 0.414 1.4
3 0.211 0.366 1.7 the sum of the two partial activities A = A1 + A2 . Each ac-
Ni − Si
4 0.167 0.333 2.0 tivity Ai can be expressed as Ai = . Accordingly, the
5 0.138 0.309 2.2 ηi ti
10 0.076 0.24 3.2 decision threshold is equal to:
20 0.041 0.183 4.5
√ √
50 0.018 0.124 7.1
∗ k 2(Ni + 1) 2σNi
100 0.0009 0.091 10.0 γi = =k . (31)
ηi ti ηi ti
484 A. Vivier and J. Aupiais

Table 4. 10 replicates (t0 = ts ), 10 negative decisions, α = β = 5%, the addition of results shows
that the sample contains in fact radioactive matter.

Replicate N0 γ Detection Nraw Nnet Nnet > γ ? Decision

limit δ

1 327 42 87 352 25 No <δ

2 301 40 84 320 19 No <δ
3 285 39 81 297 12 No <δ
4 323 42 86 334 11 No <δ
5 320 42 86 326 6 No <δ
6 308 41 84 340 32 No <δ
7 301 40 84 341 40 No <δ
8 265 38 79 287 22 No <δ
9 262 38 78 299 37 No <δ
10 322 42 86 346 24 No <δ

Cumulative 3014 128 258 3242 228 Yes 228 ± 79

total

Fig. 4. Synoptic for the addition

of several activities; case A – un-
specified partition of an object
and independent measurements
for each part. The total activity of
the initial object is the sum over
its partial activities, case B – sev-
eral independent measurements of
the same object, cumulative total
obtained by weighted average.

By reminding that the decision threshold is made by taking tion is found:

into account the distribution of the measured activity with
respect to the hypothesis H0 , i.e. a non-radioactive sample,
one gets the random variable: γ∗ = (γ1∗ )2 + (γ2∗ )2 . (34)

N2 − N1 N2 − N1
[A]H0 = [A1 ]H0 + [A2 ]H0 = + , (32) In conclusion, for n partial measurements and the corres-
η1 t1 η2 t2
ponding partial decision thresholds γi∗ associated with a risk
characterized by a null mean and a standard deviation equal α the overall decision threshold is equal to:
to:
σN2  + σN2  
σN2 2 + σN2 1 2
σ2
[A]H0 =σ 2
[A1 ]H0 +σ 2
[A2 ]H0 = + 2 2 ∗
γ = γi∗ . (35)
(η1 t1 ) 2
(η2 t2 ) 2
i
2σN2 1 2σN2 2
= + (33)
(η1 t1 )2 (η2 t2 )2
Case B: The method of collective management of the
Since the decision threshold is given by γ = k1−α σ[A]H0 , by results lies on the principle of the weighted averages. The
using Eq. 32 for both components of A, the following equa- weighted average expressed in activity is written:
Optimization of the decision threshold for single radioactive counting 485

Table 5. Replicate = 10, t0 = ts , 10 negative decisions, α = β = 5%. The cumulative total shows the presence of
133
Ba.

E (keV) Iγ (%) ε (%) Nraw Nnet N1 γ ∗ (kBq) A (kBq) A > γ ∗? Relative

uncertainty

81 36.8 6.8 390 26 364 21 10 No 211

276 7.2 6.1 329 29 300 109 66 No 173
303 18.3 5.8 345 41 304 45 39 No 124
356 62.0 4.8 424 46 378 18 15 No 123
384 8.9 4.3 403 7 396 144 18 No 808

Addition 13 16 Yes 84

Table 6. Simulation of gross alpha aqueous releases – efficiency 30%, time of counting 60 s, sample volume 100 mL. The cumulative decision
threshold and the detection limit δ are calculated according to the propagation error law.

Month Nraw N1 Nnet Released γ∗ δ Decision A Released u AR

volume (m3 ) (MBq) (MBq) > γ∗ (Bq/m3 ) activity (95% conf.
AR level)

January 1515 1492 23 1500 64 129 no 0.01 19 91

February 1526 1474 52 1200 51 102 no 0.03 35 73
March 1506 1465 41 3000 128 255 no 0.02 68 182
April 1513 1444 69 2500 106 211 no 0.04 96 151
May 1518 1545 −27 3500 153 306 no −0.02 −53 215
June 1543 1501 42 2000 86 172 no 0.02 47 123
July 1547 1492 55 1800 77 155 no 0.03 55 110
August 1507 1537 −30 2400 105 209 no −0.02 −40 147
September 1544 1468 76 2800 119 238 no 0.04 118 171
October 1586 1518 68 2000 87 173 no 0.04 76 124
November 1499 1441 58 1600 67 135 no 0.03 52 96
December 1503 1537 −34 2700 118 235 no −0.02 −51 165

Cumulative total 27 000 349 698 yes 422 496


n
Ai composition of γ -rays and the associated background allows
u2Ai
A= 
i=1
. (36) the detection of radioactive matter even if every activity de-
n
1 termined from one γ -ray does not allow to decide whether
u2Ai radioactivity is present or not. An other example given in
i=1
Table 6 concerns the management of results. Since the de-
The corresponding associated uncertainty is: cision threshold is homogeneous to a standard deviation,
 their combination should be performed using the rules of
 1
 composition for standard deviations and not, as usually en-
uA =   . (37)
 n 1 countered, be a simple addition of numbers but justified as
u2Ai
i=1 example for the addition of activity.

The total decision threshold from n partial decision thresh-

olds γi∗ , is equal to: 3.6 Gamma ray spectrometry – examples of
optimization of the Region of Interest (ROI) and

 1 background evaluation
∗ 
γ =  1 . (38)
The size of ROI in γ -spectrometry has a strong effect on
i ( i )
γ ∗ 2
the sensitivity due to two contradictory effects: – ROI too
small: γ ∗ increases as efficiency decreases; – ROI too large:
In the case where n measurements lead to comparable deci-
γ ∗ increases as background noise increases. Optimisation of
sion thresholds γ ∗ , the method of addition makes it possible
signal to noise ratio leads to choose an optimal width equal
to obtain an overall decision threshold equal to:
to 2.5 the fwhm (full width at half maximum). It is also
γ∗ possible to decrease the decision threshold by reducing the
γ∗ = √ . (39) uncertainty associated with the evaluation of the background
n
under the peak. The background noise when performing
The Eq. (36) and Eq. (39) have been applied in γ -spectro- peak treatment on a γ -spectrum is generally estimated on
metry where several γ -rays are used to determine the ac- the left side (region a, l1 channels) and on the right side (re-
tivity of radioactive isotopes. Table 5 shows that a judicious gion a , l2 channels) of the ROI. The area under the peak is
486 A. Vivier and J. Aupiais

Table 7. Expected improvement of the decision threshold by optimizing the region of interest (ROI
in γ -spectrometry).

Radionuclide Expected gain for δ Maximum gain by optimizing Corresponding factor of increase
by using all γ -peaks a the background under the peak b of the counting time c

Co 56 1.20 1.69 2.87

Fe 59 1.28 1.81 3.29
Se 75 1.37 1.93 3.74
Co 60 1.41 1.99 3.96
Kr 88 1.23 1.74 3.01
Kr 89 1.20 1.69 2.87
Zr 95 1.25 1.77 3.14
Ag 100m 1.11 1.56 2.45
Sb 125 1.30 1.84 3.38
Sb 126 1.24 1.76 3.10
I 132 1.23 1.74 3.03
Ba 133 1.38 1.94 3.78
Cs 134 1.27 1.80 3.24
I 135 1.48 2.10 4.39
Eu 152 1.58 2.23 4.99
Eu 154 1.07 1.51 2.28
Ta 182 1.21 1.71 2.94
Pb 211 1.26 1.79 3.19
Ac 227 1.71 2.42 5.83
Ac 228 1.61 2.28 5.18
Pa 231 1.80 2.55 6.48
Pa 334 m 1.61 2.27 5.17
Th 234 1.57 2.22 4.91
U 237 1.57 2.22 4.92
Pu 239 1.08 1.53 2.34
Am 241 1.20 1.70 2.88

a: Referred to a γ -peak leading to the lowest decision threshold;

b: Expecting value if a sufficient number of channels can be √ taken at the right and the left sides of the γ -peak. The
second column corresponds to the first one multiplied by 2;
c: factor of increase of the counting time required to obtain the same sensitivity as in column 1.

estimated by the surface of a trapezoid: Thus, the decision threshold falls as the number of chan-
nels used to determine the background noise under the peak
(N1 + N2 )(a − a + 1) increases. Under optimal
N̂0 = , (40) √ conditions, one can gain on each
2 peak a factor equal to 2. It will be noted that this method

l1 
l2 does not decrease the estimation of the background noise but
Ni Nj merely its fluctuations. Examples of expecting optimisation
with N1 = i=1
l1
and N2 = i=1
l2
. The associated uncertain- are given in Table 7.
N1 N2
ties are uN1 = √ and uN2 = √ , respectively. If b is the
l1 l2
width of the ROI, then the uncertainty associated with N̂0 is: 3.7 Other methods: Available informations
below the decision threshold.

Contribution of the Bayesian approach
2 N1 N2
uN̂0 = + . (41) The Bayesian treatment allows extracting more information
b l1 l2
when using all available data. By construction of the de-
Therefore, two cases are considered: cision threshold, only single information is required: the

 background counting value. Here, the background (single
b measurement N1 during t0 ) and the sample (single meas-
γ = k1−α N̂0 1 + (case l1 = l2 ) (42)
2l urement Ns during ts ) are both considered. The densities of
probability of the corresponding true means µB and µs are
l1 +l2
where l = , and
2
calculated according to the following relations:


 e−nµB (nµB )N1
 b 1 1 f(µ B |N1 ) = , (44)
 N1 + N2 + + N1 !
 2  2l1 2l2 

γ = k1−α 
2
+ b N 1 N 2 1 1 and
+ + 2+ 2
2 l1 l2 2l1 2l2
e−µs µNs s
g(µ s |Ns ) = . (45)
(case l1  = l2 ) . (43) Ns !
Optimization of the decision threshold for single radioactive counting 487

Fig. 5. Left y-axis: universal curve giving the

probability for a source to be radioactive in
function of the ratio Nnet /γ for α = 2.5%
(straight) and α = 5% (dash-dot); right y-
axis: variation of the maximum true mean
weighted by the detection limit (δ = k21−β +
2γ ) in function of Nnet /γ for γ = 2.5%
(straight) and α = 5% (dash-dot).

Since the true mean of the net countings is µnet = µ s − µ B, threshold and the detection limit for α = β = 2.5% are γ =
the density of probability is obtained by calculating the con- 27 counts and δ = 59 counts, respectively. For α = β = 5%
volution product f × g of all possible values with respect to the corresponding values are γ = 23 counts and δ = 49
the relation µnet = µs − µB : counts. Let consider a sample for which the raw number of
 count is 112 counts, i.e. Nnet = 12 counts. In the common
h(µnet |N1 , Ns ) = f(µ B |N1 ) × g(µ s |Ns )n dµB . (46) approach, the sample is considered as non radioactive since
µB ≥0
below the decision threshold but with a probability β that its
activity is at maximum equal to δ. The complete Bayesian
By replacing the corresponding term given by Eq. (44) approach developed here allows declaring whatever the risk
and Eq. (45) into Eq. (46), the distribution of µnet can (see Fig. 5) that the probability for the source to be radioac-
be determined and its integral represents the probability tive is 80% and that the corresponding maximum true mean
that the source be radioactive. This equation is similar to is below 40. This last value could be used as the detection
those proposed by Little [10] and Miller [13]. Netherveless, limit associated with the sample, on the contrary of δ which
we can not compare them since our approach is based on is associated to the system of detection.
the determination of single information, namely the back-
ground, without any additional information about the system
of detection. This situation is common, for instance in γ -
spectrometry where the background is variable from sample
4. Conclusion
to sample. We have numerically noticed that Eq. (46) can be A new complete Bayesian treatment has been applied to
satisfactorily replaced by the positive part of the Gauss dis- define the decision threshold. It leads to introduce new
tribution with a true mean µnet = 
Ns − Nn1 (µ may be negative) concepts: a standard deviation associated with the deci-
and a standard deviation σµnet = Ns + Nn1 . sion threshold where no decision can be taken within γ ±
uγ range, a discrete value of the quantile (recommended
This density of probability allows determining two quan-
use for very low counts but can be satisfactorily approxi-
tities of interest. The first one is the probability for
 a source mated by the continuous Gaussian distribution for N > 10
to be radioactive, i.e. P(µnet > 0|N1 , Ns ) = h(µ)dµ.
µnet >0 counts). Comparisons with classical approach as well as the
Usually, the acceptance of the hypothesis H0 consists in fo- Bayesian treatment in the new norm ISO 11929 show iden-
cusing on µnet = 0. Nevertheless, the calculation shows that tical results for N enough high which demonstrate a great
the probability for a source to be radioactive is still im- coherence. Differences are observed for N small due to ap-
portant just below the decision threshold. For Nnet = 0, the proximations applied by the other approaches. The correct
probability is 50%. Universal curve showing the variation of justification on the use of the Gauss distribution is numer-
this probability in function of the ratio Nγnet is given in Fig. 5. ically and quantitatively illustrated. The Bayesian treatment
The second quantity of interest is the maximum net also gives more relevant informations like the probabil-
counting µmax of the sample such as P(µnet > µmax ) < β. ity that a source be radioactive when considering the true
If you consider the quantity µmax /δ (where δ is the detec- mean of both background and sample. This concept could
tion limit such as δ = k21−β + 2γ ), its variation in function advantageously replace the common use of the decision
of Nnet /γ is linear. As example, for N1 = 100, the decision threshold.
488 A. Vivier and J. Aupiais

Annexe

Table A1. Density of probability BΦ for the centred reduced law N2 |N1 (probability to find a value below k1−α ). Distribution of N2 |N1 is discrete
as well as k1−α . The distribution is not symmetric.

N1 = 0 N1 = 1 N1 = 2 N1 = 3 N1 = 4 N1 = 5 N1 = 6 N1 = 7 N1 = 8 N1 = 9
k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ

0.000 0.5000 −0.500 0.2500 −0.817 0.1250 −1.061 0.0625 −1.265 0.0313 −1.443 0.0156 −1.604 0.0078 −1.750 0.0039 −1.886 0.0020 −2.013 0.0010
0.707 0.7500 0.000 0.5000 −0.408 0.3125 −0.707 0.1875 −0.949 0.1094 −1.155 0.0625 −1.336 0.0352 −1.500 0.0195 −1.650 0.0107 −1.789 0.0059
1.414 0.8750 0.500 0.6875 0.000 0.5000 −0.354 0.3438 −0.633 0.2266 −0.866 0.1445 −1.069 0.0898 −1.250 0.0547 −1.414 0.0327 −1.565 0.0193
2.121 0.9375 1.000 0.8125 0.4083 0.6563 0.000 0.5000 −0.316 0.3633 −0.577 0.2539 −0.802 0.1719 −1.000 0.1133 −1.179 0.0730 −1.342 0.0461
2.828 0.9688 1.500 0.8906 0.817 0.7734 0.354 0.6367 0.000 0.5000 −0.289 0.3770 −0.535 0.2744 −0.750 0.1939 −0.943 0.1334 −1.118 0.0898
3.536 0.9844 2.000 0.9375 1.225 0.8555 0.707 0.7461 0.316 0.6231 0.000 0.5000 −0.267 0.3872 −0.500 0.2905 −0.707 0.2120 −0.894 0.1509
4.243 0.9922 2.500 0.9648 1.633 0.9102 1.061 0.8281 0.633 0.7256 0.289 0.6128 0.000 0.5000 −0.250 0.3953 −0.471 0.3036 −0.671 0.2273
4.950 0.9961 3.000 0.9805 2.041 0.9453 1.414 0.8867 0.949 0.8062 0.577 0.7095 0.267 0.6047 0.000 0.5000 −0.236 0.4018 −0.447 0.3145
5.657 0.9981 3.500 0.9893 2.450 0.9673 1.768 0.9270 1.265 0.8666 0.866 0.7880 0.535 0.6964 0.250 0.5982 0.000 0.5000 −0.224 0.4073
6.364 0.9990 4.000 0.9941 2.858 0.9807 2.121 0.9539 1.581 0.9102 1.155 0.8491 0.802 0.7728 0.500 0.6855 0.236 0.5927 0.000 0.5000
7.071 0.9995 4.500 0.9968 3.266 0.9888 2.475 0.9713 1.897 0.9408 1.443 0.8949 1.069 0.8339 0.750 0.7597 0.471 0.6762 0.224 0.5881
7.778 0.9998 5.000 0.9983 3.674 0.9935 2.828 0.9824 2.214 0.9616 1.732 0.9283 1.336 0.8811 1.000 0.8204 0.707 0.7483 0.447 0.6682
8.486 0.9999 5.500 0.9991 4.083 0.9963 3.182 0.9894 2.530 0.9755 2.021 0.9519 1.604 0.9165 1.250 0.8684 0.943 0.8083 0.671 0.7383
6.000 0.9995 4.491 0.9979 3.536 0.9936 2.846 0.9846 2.309 0.9682 1.871 0.9423 1.500 0.9054 1.179 0.8569 0.894 0.7976
6.500 0.9997 4.899 0.9988 3.889 0.9962 3.162 0.9904 2.598 0.9793 2.138 0.9608 1.750 0.9331 1.414 0.8950 1.118 0.8463
7.000 0.9999 5.307 0.9993 4.243 0.9978 3.479 0.9941 2.887 0.9867 2.405 0.9738 2.000 0.9534 1.650 0.9242 1.342 0.8852
5.716 0.9996 4.596 0.9987 3.795 0.9964 3.175 0.9920 2.673 0.9827 2.250 0.9680 1.886 0.9461 1.565 0.9157
6.124 0.9998 4.950 0.9993 4.111 0.9978 3.464 0.9947 2.940 0.9887 2.500 0.9784 2.121 0.9622 1.789 0.9390
6.532 0.9999 5.303 0.9996 4.427 0.9987 3.753 0.9967 3.207 0.9927 2.750 0.9855 2.357 0.9739 2.013 0.9564
5.657 0.9998 4.743 0.9992 4.042 0.9980 3.474 0.9953 3.000 0.9904 2.593 0.9822 2.236 0.9693
6.010 0.9999 5.060 0.9995 4.330 0.9988 3.742 0.9970 3.250 0.9937 2.828 0.9879 2.460 0.9786
5.376 0.9997 4.619 0.9992 4.009 0.9981 3.500 0.9959 3.064 0.9919 2.683 0.9853
5.692 0.9998 4.908 0.9995 4.276 0.9988 3.750 0.9974 3.300 0.9947 2.907 0.9900
6.008 0.9999 5.196 0.9997 4.543 0.9993 4.000 0.9983 3.536 0.9965 3.131 0.9932
6.325 1.0000 5.485 0.9998 4.811 0.9996 4.250 0.9990 3.771 0.9977 3.354 0.9955
5.774 0.9999 5.078 0.9997 4.500 0.9993 4.007 0.9985 3.578 0.9970
5.345 0.9998 4.750 0.9996 4.243 0.9991 3.801 0.9980
5.613 0.9999 5.000 0.9998 4.478 0.9994 4.025 0.9987
5.250 0.9998 4.714 0.9996 4.249 0.9992
5.500 0.9999 4.950 0.9998 4.472 0.9995
5.186 0.9999 4.696 0.9997
4.919 0.9998
5.143 0.9999

N1 = 10 N1 = 11 N1 = 12 N1 = 13 N1 = 14 N1 = 15 N1 = 16 N1 = 17 N1 = 18 N1 = 19
k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ

−2.132 0.0000 −2.245 0.0000 −2.353 0.0000 −2.457 0.0000 −2.556 0.0000 −2.652 0.0000 −2.744 0.0000
−1.919 0.0032 −2.041 0.0017 −2.157 0.0009 −2.268 0.0000 −2.374 0.0003 −2.475 0.0001 −2.573 0.0001 −2.667 0.0000 −2.758 0.0000
−1.706 0.0112 −1.837 0.0065 −1.961 0.0037 −2.079 0.0021 −2.191 0.0012 −2.298 0.0007 −2.401 0.0004 −2.500 0.0002 −2.596 0.0001 −2.688 0.0001
−1.492 0.0287 −1.633 0.0176 −1.765 0.0106 −1.890 0.0064 −2.008 0.0038 −2.121 0.0022 −2.230 0.0013 −2.333 0.0007 −2.433 0.0004 −2.530 0.0002
−1.279 0.0592 −1.429 0.0384 −1.569 0.0245 −1.701 0.0154 −1.826 0.0096 −1.944 0.0059 −2.058 0.0036 −2.167 0.0022 −2.271 0.0013 −2.372 0.0008
−1.066 0.1051 −1.225 0.0717 −1.373 0.0481 −1.512 0.0318 −1.643 0.0207 −1.768 0.0133 −1.887 0.0085 −2.000 0.0053 −2.109 0.0033 −2.214 0.0020
−0.853 0.1662 −1.021 0.1189 −1.177 0.0835 −1.323 0.0577 −1.461 0.0392 −1.591 0.0262 −1.715 0.0173 −1.833 0.0113 −1.947 0.0073 −2.056 0.0047
−0.640 0.2403 −0.817 0.1796 −0.981 0.1316 −1.134 0.0946 −1.278 0.0669 −1.414 0.0466 −1.544 0.0320 −1.667 0.0216 −1.784 0.0145 −1.897 0.0096
−0.426 0.3238 −0.612 0.2517 −0.785 0.1917 −0.945 0.1431 −1.096 0.1050 −1.237 0.0758 −1.372 0.0539 −1.500 0.0378 −1.622 0.0261 −1.739 0.0179
−0.213 0.4119 −0.408 0.3318 −0.588 0.2617 −0.756 0.2024 −0.913 0.1537 −1.061 0.1148 −1.201 0.0843 −1.333 0.0610 −1.460 0.0436 −1.581 0.0307
0.000 0.5000 −0.204 0.4159 −0.392 0.3388 −0.567 0.2706 −0.730 0.2122 −0.884 0.1635 −1.029 0.1239 −1.167 0.0925 −1.298 0.0680 −1.423 0.0494
0.213 0.5841 0.000 0.5000 −0.196 0.4194 −0.378 0.3450 −0.548 0.2786 −0.707 0.2210 −0.858 0.1725 −1.000 0.1325 −1.136 0.1002 −1.265 0.0748
0.426 0.6612 0.204 0.5806 0.000 0.5000 −0.189 0.4225 −0.365 0.3506 −0.530 0.2858 −0.686 0.2291 −0.833 0.1808 −0.973 0.1405 −1.107 0.1077
0.640 0.7294 0.408 0.6550 0.196 0.5775 0.000 0.5000 −0.183 0.4253 −0.354 0.3555 −0.515 0.2923 −0.667 0.2366 −0.811 0.1885 −0.949 0.1481
0.853 0.7878 0.612 0.7214 0.392 0.6495 0.189 0.5747 0.000 0.5000 −0.177 0.4278 −0.343 0.3601 −0.500 0.2983 −0.649 0.2434 −0.791 0.1958
1.066 0.8365 0.817 0.7790 0.588 0.7142 0.378 0.6445 0.183 0.5722 0.000 0.5000 −0.172 0.4300 −0.333 0.3642 −0.487 0.3038 −0.633 0.2498
1.279 0.8761 1.021 0.8275 0.785 0.7709 0.567 0.7077 0.365 0.6400 0.177 0.5700 0.000 0.5000 −0.167 0.4321 −0.324 0.3679 −0.474 0.3089
1.492 0.9075 1.225 0.8675 0.981 0.8192 0.756 0.7634 0.548 0.7017 0.354 0.6358 0.172 0.5679 0.000 0.5000 −0.162 0.4340 −0.316 0.3714
1.706 0.9320 1.429 0.8998 1.177 0.8595 0.945 0.8115 0.730 0.7566 0.530 0.6962 0.343 0.6321 0.167 0.5660 0.000 0.500 −0.158 0.4357
1.919 0.9506 1.633 0.9252 1.373 0.8923 1.134 0.8519 0.913 0.8042 0.707 0.7502 0.515 0.6911 0.333 0.6286 0.162 0.5643 0.000 0.5000
2.132 0.9646 1.837 0.9449 1.569 0.9186 1.323 0.8853 1.096 0.8448 0.884 0.7975 0.686 0.7443 0.500 0.6865 0.324 0.6254 0.158 0.5627
2.345 0.9750 2.041 0.9599 1.765 0.9393 1.512 0.9123 1.278 0.8785 1.061 0.8380 0.858 0.7912 0.667 0.7388 0.487 0.6821 0.316 0.6224
2.558 0.9825 2.245 0.9712 1.961 0.9552 1.701 0.9338 1.461 0.9061 1.237 0.8721 1.029 0.8316 0.833 0.7852 0.649 0.7337 0.474 0.6780
2.772 0.9879 2.450 0.9795 2.157 0.9674 1.890 0.9506 1.643 0.9284 1.414 0.9002 1.201 0.8659 1.000 0.8256 0.811 0.7796 0.633 0.7288
2.985 0.9917 2.654 0.9856 2.353 0.9765 2.079 0.9635 1.826 0.9459 1.591 0.9231 1.372 0.8945 1.167 0.8600 0.973 0.8198 0.791 0.7743
3.198 0.9943 2.858 0.9900 2.550 0.9832 2.268 0.9734 2.008 0.9597 1.768 0.9414 1.544 0.9179 1.333 0.8890 1.136 0.8544 0.949 0.8144
3.411 0.9962 3.062 0.9931 2.746 0.9882 2.457 0.9808 2.191 0.9702 1.945 0.9558 1.715 0.9369 1.500 0.9129 1.298 0.8837 1.107 0.8490
3.624 0.9975 3.266 0.9953 2.942 0.9917 2.646 0.9862 2.374 0.9782 2.121 0.9670 1.887 0.9519 1.667 0.9324 1.460 0.9080 1.265 0.8785
3.838 0.9983 3.470 0.9968 3.138 0.9943 2.835 0.9902 2.556 0.9842 2.298 0.9756 2.058 0.9638 1.833 0.9481 1.622 0.9281 1.423 0.9033
4.051 0.9989 3.674 0.9978 3.334 0.9960 3.024 0.9931 2.739 0.9887 2.475 0.9822 2.230 0.9730 2.000 0.9605 1.784 0.9443 1.581 0.9238
4.264 0.9993 3.878 0.9986 3.530 0.9973 3.213 0.9952 2.921 0.9920 2.652 0.9871 2.401 0.9800 2.167 0.9703 1.947 0.9573 1.739 0.9405
4.477 0.9995 4.083 0.9990 3.726 0.9982 3.402 0.9967 3.104 0.9943 2.828 0.9907 2.573 0.9854 2.333 0.9778 2.109 0.9676 1.897 0.9540
4.690 0.9997 4.287 0.9994 3.922 0.9988 3.591 0.9977 3.286 0.9960 3.005 0.9934 2.744 0.9894 2.500 0.9836 2.271 0.9756 2.056 0.9648
4.904 0.9998 4.491 0.9996 4.118 0.9992 3.780 0.9985 3.469 0.9972 3.182 0.9953 2.916 0.9923 2.667 0.9880 2.433 0.9818 2.214 0.9733
5.117 0.9998 4.695 0.9997 4.315 0.9995 3.969 0.9990 3.652 0.9981 3.359 0.9967 3.087 0.9945 2.833 0.9912 2.596 0.9865 2.372 0.9799
5.330 0.9999 4.899 0.9998 4.511 0.9996 4.158 0.9993 3.834 0.9987 3.536 0.9977 3.259 0.9961 3.000 0.9937 2.758 0.9901 2.530 0.9850
5.543 1.0000 5.103 0.9999 4.707 0.9998 4.347 0.9995 4.017 0.9991 3.712 0.9984 3.430 0.9973 3.167 0.9955 2.920 0.9928 2.688 0.9889
4.903 0.9999 4.536 0.9997 4.199 0.9994 3.889 0.9989 3.602 0.9981 3.333 0.9968 3.082 0.9948 2.846 0.9918
5.099 0.9999 4.725 0.9998 4.382 0.9996 4.066 0.9993 3.773 0.9987 3.500 0.9977 3.244 0.9963 3.004 0.9940
4.914 0.9999 4.564 0.9997 4.243 0.9995 3.945 0.9991 3.667 0.9984 3.407 0.9973 3.162 0.9957
Optimization of the decision threshold for single radioactive counting 489

4.747 0.9998 4.419 0.9997 4.116 0.9994 3.833 0.9989 3.569 0.9981 3.320 0.9969
4.930 0.9999 4.596 0.9998 4.288 0.9996 4.000 0.9992 3.731 0.9987 3.479 0.9978
4.773 0.9999 4.459 0.9997 4.167 0.9995 3.893 0.9991 3.637 0.9984
4.631 0.9998 4.333 0.9996 4.056 0.9994 3.795 0.9989
4.802 0.9999 4.500 0.9998 4.218 0.9996 3.953 0.9992
4.667 0.9998 4.380 0.9997 4.111 0.9995
4.833 0.9999 4.542 0.9998 4.269 0.9996
4.704 0.9999 4.427 0.9997
4.585 0.9998
4.743 0.9999

N1 = 20 N1 = 21 N1 = 22 N1 = 23 N1 = 24 N1 = 25 N1 = 26 N1 = 27 N1 = 28 N1 = 29
k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ

−2.778 0.0000 −2.864 0.0000

−2.623 0.0001 −2.714 0.0001 −2.801 0.0000 −2.887 0.0000 −2.970 0.0000
−2.469 0.0005 −2.563 0.0003 −2.654 0.0002 −2.742 0.0001 −2.828 0.0001 −2.912 0.0000 −2.994 0.0000
−2.315 0.0013 −2.412 0.0008 −2.507 0.0005 −2.598 0.0003 −2.687 0.0002 −2.774 0.0001 −2.858 0.0001 −2.940 0.0000 −3.020 0.0000
−2.160 0.0030 −2.261 0.0019 −2.359 0.0012 −2.454 0.0007 −2.546 0.0004 −2.635 0.0003 −2.722 0.0002 −2.806 0.0001 −2.889 0.0001 −2.969 0.0000
−2.006 0.0063 −2.111 0.0041 −2.212 0.0026 −2.309 0.0017 −2.404 0.0011 −2.496 0.0007 −2.586 0.0004 −2.673 0.0003 −2.757 0.0002 −2.840 0.0001
−1.852 0.0121 −1.960 0.0081 −2.064 0.0053 −2.165 0.0035 −2.263 0.0023 −2.358 0.0015 −2.450 0.0009 −2.539 0.0006 −2.626 0.0004 −2.711 0.0002
−1.697 0.0214 −1.809 0.0147 −1.917 0.0100 −2.021 0.0068 −2.121 0.0045 −2.219 0.0030 −2.313 0.0020 −2.405 0.0013 −2.495 0.0008 −2.582 0.0005
−1.543 0.0354 −1.658 0.0251 −1.769 0.0175 −1.876 0.0122 −1.980 0.0083 −2.080 0.0057 −2.177 0.0038 −2.272 0.0026 −2.364 0.0017 −2.453 0.0011
−1.389 0.0551 −1.508 0.0401 −1.622 0.0288 −1.732 0.0205 −1.839 0.0144 −1.942 0.0100 −2.041 0.0069 −2.138 0.0047 −2.232 0.0032 −2.324 0.0022
−1.234 0.0814 −1.357 0.0607 −1.474 0.0448 −1.588 0.0326 −1.697 0.0235 −1.803 0.0168 −1.905 0.0119 −2.005 0.0083 −2.101 0.0058 −2.195 0.0040
−1.080 0.1147 −1.206 0.0877 −1.327 0.0663 −1.443 0.0494 −1.556 0.0365 −1.664 0.0266 −1.769 0.0192 −1.871 0.0138 −1.970 0.0098 −2.066 0.0069
−0.926 0.1553 −1.055 0.1215 −1.180 0.0939 −1.299 0.0717 −1.414 0.0541 −1.525 0.0404 −1.633 0.0298 −1.737 0.0218 −1.838 0.0158 −1.937 0.0113
−0.772 0.2025 −0.905 0.1620 −1.032 0.1279 −1.155 0.0998 −1.273 0.0769 −1.387 0.0586 −1.497 0.0442 −1.604 0.0330 −1.707 0.0244 −1.807 0.0179
−0.617 0.2557 −0.754 0.2089 −0.885 0.1684 −1.010 0.1341 −1.131 0.1055 −1.248 0.0821 −1.361 0.0631 −1.470 0.0481 −1.576 0.0362 −1.678 0.0270
−0.463 0.3136 −0.603 0.2612 −0.737 0.2148 −0.866 0.1744 −0.990 0.1400 −1.109 0.1110 −1.225 0.0871 −1.336 0.0676 −1.444 0.0519 −1.549 0.0395
−0.309 0.3746 −0.452 0.3179 −0.590 0.2664 −0.722 0.2204 −0.849 0.1802 −0.971 0.1456 −1.089 0.1164 −1.203 0.0920 −1.313 0.0719 −1.420 0.0557
−0.154 0.4373 −0.302 0.3776 −0.442 0.3220 −0.577 0.2712 −0.707 0.2257 −0.832 0.1857 −0.953 0.1510 −1.069 0.1215 −1.182 0.0967 −1.291 0.0762
0.000 0.5000 −0.151 0.4388 −0.295 0.3804 −0.433 0.3258 −0.566 0.2757 −0.693 0.2307 −0.817 0.1909 −0.935 0.1562 −1.051 0.1264 −1.162 0.1013
0.154 0.5612 0.000 0.5000 −0.147 0.4402 −0.289 0.3830 −0.424 0.3294 −0.555 0.2800 −0.680 0.2354 −0.802 0.1958 −0.919 0.1611 −1.033 0.1312
0.309 0.6196 0.151 0.5598 0.000 0.5000 −0.144 0.4415 −0.283 0.3854 −0.416 0.3327 −0.544 0.2841 −0.668 0.2399 −0.788 0.2005 −0.904 0.1659
0.463 0.6742 0.302 0.6170 0.147 0.5585 0.000 0.5000 −0.141 0.4427 −0.277 0.3877 −0.408 0.3359 −0.535 0.2879 −0.657 0.2442 −0.775 0.2051
0.617 0.7243 0.452 0.6706 0.295 0.6146 0.144 0.5573 0.000 0.5000 −0.139 0.4439 −0.272 0.3899 −0.401 0.3389 −0.525 0.2916 −0.646 0.2483
0.772 0.7693 0.603 0.7200 0.442 0.6673 0.288 0.6123 0.141 0.5561 0.000 0.5000 −0.136 0.4449 −0.267 0.3919 −0.394 0.3417 −0.516 0.2950
0.926 0.8092 0.754 0.7646 0.590 0.7159 0.433 0.6641 0.283 0.6101 0.139 0.5551 0.000 0.5000 −0.134 0.4460 −0.263 0.3939 −0.387 0.3444
1.080 0.8438 0.905 0.8042 0.737 0.7601 0.577 0.7121 0.424 0.6611 0.277 0.6081 0.136 0.5540 0.000 0.5000 −0.131 0.4469 −0.258 0.3957
1.234 0.8736 1.055 0.8389 0.885 0.7995 0.722 0.7558 0.566 0.7084 0.416 0.6583 0.272 0.6062 0.134 0.5531 0.000 0.5000 −0.129 0.4478
1.389 0.8987 1.206 0.8688 1.032 0.8341 0.866 0.7950 0.707 0.7517 0.555 0.7050 0.408 0.6556 0.267 0.6043 0.131 0.5522 0.000 0.5000
1.543 0.9196 1.357 0.8942 1.180 0.8642 1.010 0.8296 0.849 0.7906 0.693 0.7478 0.544 0.7017 0.401 0.6530 0.263 0.6026 0.129 0.5513
1.697 0.9368 1.508 0.9155 1.327 0.8898 1.155 0.8597 0.990 0.8252 0.832 0.7865 0.680 0.7441 0.535 0.6985 0.394 0.6506 0.258 0.6009
1.852 0.9508 1.658 0.9332 1.474 0.9115 1.299 0.8856 1.131 0.8554 0.971 0.8209 0.817 0.7825 0.668 0.7405 0.525 0.6955 0.387 0.6482
2.006 0.9621 1.809 0.9476 1.622 0.9296 1.443 0.9076 1.273 0.8815 1.109 0.8512 0.953 0.8169 0.802 0.7787 0.657 0.7371 0.516 0.6927
2.160 0.9710 1.960 0.9593 1.769 0.9444 1.588 0.9260 1.414 0.9037 1.248 0.8775 1.089 0.8472 0.935 0.8129 0.788 0.7750 0.646 0.7339
2.315 0.9780 2.111 0.9686 1.917 0.9565 1.732 0.9413 1.556 0.9225 1.387 0.9000 1.225 0.8736 1.069 0.8433 0.919 0.8091 0.775 0.7715
2.469 0.9834 2.261 0.9760 2.064 0.9663 1.876 0.9538 1.697 0.9381 1.525 0.9191 1.361 0.8963 1.203 0.8698 1.051 0.8395 0.904 0.8055
2.623 0.9876 2.412 0.9818 2.212 0.9741 2.021 0.9639 1.839 0.9510 1.664 0.9350 1.497 0.9157 1.336 0.8927 1.182 0.8661 1.033 0.8358
2.778 0.9908 2.563 0.9863 2.359 0.9802 2.165 0.9721 1.980 0.9615 1.803 0.9483 1.633 0.9320 1.470 0.9124 1.313 0.8892 1.162 0.8625
2.932 0.9933 2.714 0.9898 2.507 0.9850 2.309 0.9785 2.121 0.9700 1.942 0.9592 1.769 0.9456 1.604 0.9290 1.444 0.9091 1.291 0.8858
3.086 0.9951 2.864 0.9924 2.654 0.9887 2.454 0.9836 2.263 0.9768 2.080 0.9650 1.905 0.9568 1.737 0.9429 1.576 0.9260 1.420 0.9059
3.240 0.9964 3.015 0.9944 2.801 0.9916 2.598 0.9876 2.404 0.9822 2.219 0.9751 2.041 0.9659 1.871 0.9544 1.707 0.9402 1.549 0.9230
3.395 0.9974 3.166 0.9959 2.949 0.9938 2.742 0.9907 2.546 0.9864 2.358 0.9808 2.177 0.9734 2.005 0.9639 1.838 0.9520 1.678 0.9375
3.549 0.9982 3.317 0.9970 3.096 0.9954 2.887 0.9930 2.687 0.9897 2.496 0.9853 2.313 0.9793 2.138 0.9716 1.970 0.9618 1.807 0.9497
3.703 0.9987 3.467 0.9979 3.244 0.9966 3.031 0.9948 2.828 0.9923 2.635 0.9888 2.450 0.9840 2.272 0.9778 2.101 0.9698 1.937 0.9598
3.858 0.9991 3.618 0.9985 3.391 0.9976 3.175 0.9962 2.970 0.9942 2.774 0.9915 2.586 0.9878 2.405 0.9828 2.232 0.9763 2.066 0.9680
4.012 0.9994 3.769 0.9989 3.539 0.9982 3.320 0.9972 3.111 0.9957 2.912 0.9936 2.722 0.9907 2.539 0.9867 2.364 0.9815 2.195 0.9748
4.166 0.9995 3.920 0.9992 3.686 0.9987 3.464 0.9980 3.253 0.9969 3.051 0.9952 2.858 0.9930 2.673 0.9899 2.495 0.9857 2.324 0.9802
4.321 0.9997 4.070 0.9995 3.834 0.9991 3.608 0.9985 3.394 0.9977 3.190 0.9964 2.994 0.9947 2.806 0.9923 2.626 0.9890 2.453 0.9846
4.475 0.9998 4.221 0.9996 3.981 0.9994 3.753 0.9989 3.536 0.9983 3.328 0.9974 3.130 0.9961 2.940 0.9942 2.757 0.9916 2.582 0.9881
4.629 0.9999 4.372 0.9997 4.128 0.9995 3.897 0.9992 3.677 0.9988 3.467 0.9981 3.266 0.9971 3.074 0.9956 2.889 0.9936 2.711 0.9908
4.523 0.9998 4.276 0.9997 4.042 0.9995 3.818 0.9991 3.606 0.9986 3.402 0.9978 3.207 0.9967 3.021 0.9952 2.840 0.9930
4.673 0.9999 4.423 0.9998 4.186 0.9996 3.960 0.9994 3.744 0.9990 3.538 0.9984 3.341 0.9976 3.151 0.9964 2.969 0.9947
4.571 0.9998 4.330 0.9997 4.101 0.9996 3.883 0.9993 3.674 0.9988 3.474 0.9982 3.283 0.9973 3.098 0.9960
4.718 0.9999 4.475 0.9998 4.243 0.9997 4.022 0.9995 3.810 0.9992 3.608 0.9987 3.414 0.9980 3.228 0.9970
4.619 0.9999 4.384 0.9998 4.160 0.9996 3.946 0.9994 3.742 0.9990 3.545 0.9985 3.357 0.9977
4.526 0.9998 4.299 0.9997 4.083 0.9996 3.875 0.9993 3.677 0.9989 3.486 0.9983
4.667 0.9999 4.438 0.9998 4.219 0.9997 4.009 0.9995 3.808 0.9992 3.615 0.9988
4.576 0.9999 4.355 0.9998 4.143 0.9996 3.939 0.9994 3.744 0.9991
4.491 0.9998 4.276 0.9997 4.071 0.9996 3.873 0.9993
4.627 0.9999 4.410 0.9998 4.202 0.9997 4.002 0.9995
4.543 0.9999 4.333 0.9998 4.131 0.9997
4.464 0.9998 4.260 0.9998
4.596 0.9999 4.389 0.9998
4.519 0.9999

N1 = 30 N1 = 35 N1 = 40 N1 = 45 N1 = 50 N1 = 55 N1 = 60 N1 = 65 N1 = 70
k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ k1−α BΦ

−3.048 0.0000
−2.921 0.0001
−2.794 0.0001
−2.667 0.0003 −3.064 0.0000
−2.540 0.0007 −2.946 0.0001
−2.413 0.0014 −2.828 0.0002
−2.286 0.0027 −2.711 0.0004 −3.092 0.0000
490 A. Vivier and J. Aupiais

−2.159 0.0048 −2.593 0.0007 −2.982 0.0001

−2.032 0.0081 −2.475 0.0013 −2.871 0.0002
−1.905 0.0130 −2.357 0.0023 −2.761 0.0003 −3.128 0.0000
−1.778 0.0200 −2.239 0.0039 −2.650 0.0006 −3.024 0.0001
−1.651 0.0297 −2.121 0.0063 −2.540 0.0011 −2.919 0.0002
−1.524 0.0427 −2.004 0.0099 −2.430 0.0019 −2.815 0.0003 −3.169 0.0000
−1.397 0.0595 −1.886 0.0150 −2.319 0.0031 −2.711 0.0005 −3.070 0.0001
−1.270 0.0804 −1.768 0.0220 −2.209 0.0049 −2.606 0.0009 −2.970 0.0002
−1.143 0.1058 −1.650 0.0314 −2.098 0.0076 −2.502 0.0015 −2.871 0.0003 −3.213 0.0000
−1.016 0.1358 −1.532 0.0435 −1.988 0.0113 −2.398 0.0025 −2.772 0.0005 −3.118 0.0001
−0.889 0.1705 −1.414 0.0587 −1.877 0.0164 −2.294 0.0038 −2.673 0.0008 −3.024 0.0001
−0.762 0.2094 −1.296 0.0775 −1.767 0.0232 −2.189 0.0058 −2.574 0.0012 −2.929 0.0002 −3.259 0.0000
−0.635 0.2522 −1.179 0.1000 −1.657 0.0320 −2.085 0.0085 −2.475 0.0019 −2.835 0.0004 −3.169 0.0000
−0.508 0.2983 −1.061 0.1264 −1.546 0.0432 −1.981 0.0122 −2.376 0.0029 −2.740 0.0006 −3.078 0.0001
−0.381 0.3470 −0.943 0.1568 −1.436 0.0571 −1.877 0.0172 −2.277 0.0044 −2.646 0.0010 −2.988 0.0002
−0.254 0.3974 −0.825 0.1909 −1.325 0.0740 −1.772 0.0237 −2.178 0.0064 −2.551 0.0015 −2.897 0.0003 −3.220 0.0000
−0.127 0.4487 −0.707 0.2285 −1.215 0.0941 −1.668 0.0320 −2.079 0.0092 −2.457 0.0023 −2.807 0.0005 −3.133 0.0001
0.000 0.5000 −0.589 0.2693 −1.104 0.1176 −1.564 0.0423 −1.980 0.0128 −2.362 0.0033 −2.716 0.0008 −3.046 0.0002
0.127 0.5505 −0.471 0.3127 −0.994 0.1444 −1.460 0.0550 −1.881 0.0176 −2.268 0.0048 −2.626 0.0012 −2.959 0.0002 −3.273 0.0000
0.254 0.5994 −0.354 0.3582 −0.884 0.1746 −1.355 0.0703 −1.782 0.0238 −2.173 0.0069 −2.535 0.0017 −2.872 0.0004 −3.189 0.0000
0.381 0.6460 −0.236 0.4050 −0.773 0.2080 −1.251 0.0883 −1.683 0.0315 −2.079 0.0096 −2.445 0.0025 −2.785 0.0006 −3.105 0.0001
0.508 0.6899 −0.118 0.4525 −0.663 0.2443 −1.147 0.1093 −1.584 0.0410 −1.984 0.0132 −2.354 0.0037 −2.698 0.0009 −3.021 0.0002
0.635 0.7307 0.000 0.5000 −0.552 0.2833 −1.043 0.1332 −1.485 0.0526 −1.890 0.0177 −2.263 0.0052 −2.611 0.0013 −2.937 0.0003
0.762 0.7681 0.118 0.5469 −0.442 0.3244 −0.938 0.1602 −1.386 0.0665 −1.795 0.0235 −2.173 0.0072 −2.524 0.0019 −2.853 0.0005
0.889 0.8019 0.236 0.5924 −0.331 0.3672 −0.834 0.1900 −1.287 0.0827 −1.701 0.0307 −2.082 0.0099 −2.437 0.0028 −2.769 0.0007
1.016 0.8322 0.354 0.6362 −0.221 0.4111 −0.730 0.2226 −1.188 0.1016 −1.606 0.0395 −1.992 0.0133 −2.350 0.0039 −2.685 0.0010
1.143 0.8590 0.471 0.6778 −0.110 0.4555 −0.626 0.2577 −1.089 0.1231 −1.512 0.0501 −1.901 0.0176 −2.263 0.0054 −2.602 0.0015
1.270 0.8825 0.589 0.7167 0.000 0.5000 −0.521 0.2950 −0.990 0.1472 −1.417 0.0627 −1.811 0.0230 −2.176 0.0074 −2.518 0.0021
1.397 0.9028 0.707 0.7528 0.110 0.5439 −0.417 0.3341 −0.891 0.1741 −1.323 0.0774 −1.720 0.0297 −2.089 0.0100 −2.434 0.0030
1.524 0.9201 0.825 0.7859 0.221 0.5868 −0.313 0.3747 −0.792 0.2035 −1.228 0.0944 −1.630 0.0378 −2.002 0.0132 −2.350 0.0041
1.651 0.9349 0.943 0.8159 0.331 0.6282 −0.209 0.4161 −0.693 0.2353 −1.134 0.1138 −1.539 0.0475 −1.915 0.0173 −2.266 0.0056
1.778 0.9473 1.061 0.8429 0.442 0.6677 −0.104 0.4581 −0.594 0.2692 −1.039 0.1356 −1.449 0.0590 −1.828 0.0224 −2.182 0.0075
1.905 0.9577 1.179 0.8668 0.552 0.7050 0.000 0.5000 −0.495 0.3050 −0.945 0.1599 −1.358 0.0724 −1.741 0.0286 −2.098 0.0099
2.032 0.9662 1.296 0.8879 0.663 0.7399 0.104 0.5415 −0.396 0.3424 −0.850 0.1865 −1.268 0.0878 −1.654 0.0361 −2.014 0.0130
2.159 0.9732 1.414 0.9063 0.773 0.7721 0.209 0.5821 −0.297 0.3810 −0.756 0.2153 −1.177 0.1054 −1.567 0.0450 −1.930 0.0169
2.286 0.9789 1.532 0.9222 0.884 0.8017 0.313 0.6214 −0.198 0.4204 −0.661 0.2463 −1.086 0.1251 −1.480 0.0555 −1.846 0.0216
2.413 0.9835 1.650 0.9358 0.994 0.8286 0.417 0.6591 −0.099 0.4602 −0.567 0.2792 −0.996 0.1471 −1.393 0.0677 −1.762 0.0274
2.540 0.9872 1.768 0.9474 1.104 0.8528 0.521 0.6950 0.000 0.5000 −0.473 0.3137 −0.905 0.1713 −1.306 0.0817 −1.678 0.0343
2.667 0.9901 1.886 0.9572 1.215 0.8743 0.626 0.7287 0.099 0.5394 −0.378 0.3496 −0.815 0.1976 −1.219 0.0977 −1.594 0.0425
2.794 0.9924 2.004 0.9653 1.325 0.8934 0.730 0.7602 0.198 0.5781 −0.284 0.3865 −0.724 0.2259 −1.132 0.1156 −1.511 0.0521
2.921 0.9942 2.121 0.9721 1.436 0.9102 0.834 0.7892 0.297 0.6156 −0.189 0.4241 −0.634 0.2561 −1.045 0.1356 −1.427 0.0633
3.048 0.9956 2.239 0.9777 1.546 0.9247 0.938 0.8159 0.396 0.6518 −0.095 0.4621 −0.543 0.2880 −0.957 0.1577 −1.343 0.0761
3.175 0.9967 2.357 0.9823 1.657 0.9373 1.043 0.8401 0.495 0.6863 0.000 0.5000 −0.453 0.3213 −0.870 0.1817 −1.259 0.0906
3.302 0.9975 2.475 0.9860 1.767 0.9481 1.147 0.8620 0.594 0.7190 0.095 0.5376 −0.362 0.3559 −0.783 0.2076 −1.175 0.1070
3.429 0.9981 2.593 0.9890 1.877 0.9573 1.251 0.8816 0.693 0.7496 0.189 0.5746 −0.272 0.3913 −0.696 0.2354 −1.091 0.1252
3.556 0.9986 2.711 0.9914 1.988 0.9651 1.355 0.8989 0.792 0.7782 0.284 0.6106 −0.181 0.4273 −0.609 0.2649 −1.007 0.1453
3.683 0.9990 2.828 0.9933 2.098 0.9716 1.460 0.9142 0.891 0.8045 0.378 0.6453 −0.091 0.4637 −0.522 0.2958 −0.923 0.1673
3.810 0.9992 2.946 0.9948 2.209 0.9770 1.564 0.9276 0.990 0.8287 0.473 0.6787 0.000 0.5000 −0.435 0.3281 −0.839 0.1912
3.937 0.9994 3.064 0.9960 2.319 0.9815 1.668 0.9392 1.089 0.8507 0.567 0.7104 0.091 0.5360 −0.348 0.3614 −0.755 0.2168
4.064 0.9996 3.182 0.9970 2.430 0.9852 1.772 0.9493 1.188 0.8706 0.661 0.7403 0.181 0.5715 −0.261 0.3955 −0.671 0.2440
4.191 0.9997 3.300 0.9977 2.540 0.9882 1.877 0.9579 1.287 0.8884 0.756 0.7683 0.272 0.6061 −0.174 0.4302 −0.587 0.2728
4.318 0.9998 3.418 0.9982 2.650 0.9906 1.981 0.9652 1.386 0.9043 0.850 0.7943 0.362 0.6397 −0.087 0.4651 −0.504 0.3029
4.445 0.9998 3.536 0.9987 2.761 0.9926 2.085 0.9714 1.485 0.9183 0.945 0.8183 0.453 0.6719 0.000 0.5000 −0.420 0.3341
4.572 0.9999 3.653 0.9990 2.871 0.9942 2.189 0.9766 1.584 0.9306 1.039 0.8404 0.543 0.7027 0.087 0.5347 −0.336 0.3663
3.771 0.9993 2.982 0.9955 2.294 0.9810 1.683 0.9414 1.134 0.8604 0.634 0.7318 0.174 0.5688 −0.252 0.3993
3.889 0.9995 3.092 0.9965 2.398 0.9846 1.782 0.9507 1.228 0.8785 0.724 0.7593 0.261 0.6022 −0.168 0.4327
4.007 0.9996 3.203 0.9973 2.502 0.9876 1.881 0.9588 1.323 0.8948 0.815 0.7850 0.348 0.6346 −0.084 0.4663
4.125 0.9997 3.313 0.9979 2.606 0.9901 1.980 0.9657 1.417 0.9094 0.905 0.8088 0.435 0.6659 0.000 0.5000
4.243 0.9998 3.423 0.9984 2.711 0.9921 2.079 0.9716 1.512 0.9223 0.996 0.8308 0.522 0.6958 0.084 0.5334
4.361 0.9998 3.534 0.9988 2.815 0.9937 2.178 0.9766 1.606 0.9337 1.086 0.8510 0.609 0.7243 0.168 0.5664
4.478 0.9999 3.644 0.9991 2.919 0.9950 2.277 0.9808 1.701 0.9437 1.177 0.8693 0.696 0.7512 0.252 0.5987
3.755 0.9993 3.024 0.9961 2.376 0.9843 1.795 0.9524 1.268 0.8860 0.783 0.7765 0.336 0.6301
3.865 0.9995 3.128 0.9969 2.475 0.9872 1.890 0.9599 1.358 0.9009 0.870 0.8001 0.420 0.6604
3.976 0.9996 3.232 0.9976 2.574 0.9897 1.984 0.9664 1.449 0.9143 0.957 0.8220 0.504 0.6896
4.086 0.9997 3.336 0.9981 2.673 0.9917 2.079 0.9720 1.539 0.9262 1.045 0.8422 0.587 0.7174
4.196 0.9998 3.441 0.9986 2.772 0.9933 2.173 0.9767 1.630 0.9367 1.132 0.8607 0.671 0.7438
4.307 0.9998 3.545 0.9989 2.871 0.9947 2.268 0.9808 1.720 0.9460 1.219 0.8776 0.755 0.7687
4.417 0.9999 3.649 0.9992 2.970 0.9957 2.362 0.9842 1.811 0.9541 1.306 0.8928 0.839 0.7921
3.753 0.9994 3.070 0.9966 2.457 0.9870 1.901 0.9612 1.393 0.9066 0.923 0.8138
3.858 0.9995 3.169 0.9973 2.551 0.9894 1.992 0.9673 1.480 0.9190 1.007 0.8340
3.962 0.9996 3.268 0.9979 2.646 0.9914 2.082 0.9730 1.567 0.9300 1.091 0.8526
4.066 0.9997 3.367 0.9984 2.740 0.9930 2.173 0.9771 1.654 0.9397 1.175 0.8696
4.170 0.9998 3.466 0.9987 2.835 0.9944 2.263 0.9809 1.741 0.9484 1.259 0.8852
4.275 0.9998 3.565 0.9990 2.929 0.9955 2.354 0.9842 1.828 0.9559 1.343 0.8993
4.379 0.9999 3.664 0.9992 3.024 0.9964 2.445 0.9869 1.915 0.9625 1.427 0.9120
3.763 0.9994 3.118 0.9971 2.535 0.9893 2.002 0.9683 1.511 0.9234
3.862 0.9995 3.213 0.9977 2.626 0.9912 2.089 0.9732 1.594 0.9336
3.961 0.9997 3.307 0.9982 2.716 0.9928 2.176 0.9775 1.678 0.9427
4.060 0.9997 3.402 0.9986 2.807 0.9942 2.263 0.9812 1.762 0.9507
4.159 0.9998 3.496 0.9989 2.897 0.9953 2.350 0.9843 1.846 0.9577
4.258 0.9998 3.591 0.9991 2.988 0.9962 2.437 0.9870 1.930 0.9639
3.685 0.9993 3.078 0.9969 2.524 0.9892 2.014 0.9693
3.780 0.9995 3.169 0.9975 2.611 0.9911 2.098 0.9740
3.874 0.9996 3.259 0.9980 2.698 0.9927 2.182 0.9781
3.969 0.9997 3.350 0.9984 2.785 0.9940 2.266 0.9815
4.063 0.9997 3.440 0.9988 2.872 0.9951 2.350 0.9845
4.158 0.9998 3.531 0.9990 2.959 0.9960 2.434 0.9871
4.252 0.9999 3.621 0.9992 3.046 0.9968 2.518 0.9893
Optimization of the decision threshold for single radioactive counting 491

3.712 0.9994 3.133 0.9974

3.803 0.9995 3.220 0.9979
3.893 0.9996 3.308 0.9983
3.984 0.9997 3.395 0.9987
4.074 0.9998 3.482 0.9989
4.165 0.9998
4.255 0.9999

Table A2. Error of the first kind α and the corresponding quantile k1−α for the reduced law Nnet |N1 of variance 2(N1 + 1).

N0 = 0 N0 = 1 N0 = 2 N0 = 3 N0 = 4 N0 = 5 N0 = 6 N0 = 7 N0 = 8 N0 = 9
k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α

0.000 0.7236 0.000 0.6342 0.000 0.5984 0.000 0.5805 0.000 0.5698 0.000 0.5625 0.000 0.5571 0.000 0.5529 0.000 0.5495 0.000 0.5467
0.707 0.2764 0.500 0.3658 0.408 0.4016 0.354 0.4195 0.316 0.4302 0.289 0.4375 0.267 0.4429 0.250 0.4471 0.236 0.4505 0.224 0.4533
1.414 0.1056 1.000 0.1870 0.817 0.2406 0.707 0.2764 0.633 0.3014 0.577 0.3199 0.535 0.3341 0.500 0.3455 0.471 0.3548 0.447 0.3626
2.121 0.0403 1.500 0.0894 1.225 0.1333 1.061 0.1691 0.949 0.1977 0.866 0.2207 0.802 0.2396 0.750 0.2552 0.707 0.2685 0.671 0.2798
2.828 0.0154 2.000 0.0411 1.633 0.0698 1.414 0.0975 1.265 0.1226 1.155 0.1446 1.069 0.1638 1.000 0.1806 0.943 0.1954 0.894 0.2084
3.536 0.0059 2.500 0.0183 2.041 0.0351 1.768 0.0537 1.581 0.0725 1.443 0.0905 1.336 0.1073 1.250 0.1228 1.179 0.1370 1.118 0.1499
4.243 0.0023 3.000 0.0080 2.450 0.0170 2.121 0.0284 1.897 0.0412 1.732 0.0544 1.604 0.0677 1.500 0.0805 1.414 0.0928 1.342 0.1044
4.950 0.0009 3.500 0.0034 2.858 0.0081 2.475 0.0146 2.214 0.0226 2.021 0.0316 1.871 0.0412 1.750 0.0510 1.650 0.0609 1.565 0.0705
5.657 0.0003 4.000 0.0015 3.266 0.0038 2.828 0.0073 2.530 0.0121 2.309 0.0178 2.138 0.0243 2.000 0.0314 1.886 0.0388 1.789 0.0463
6.364 0.0001 4.500 0.0006 3.674 0.0017 3.182 0.0036 2.846 0.0063 2.598 0.0098 2.405 0.0140 2.250 0.0188 2.121 0.0240 2.013 0.0296
7.071 0.0000 5.000 0.0003 4.083 0.0008 3.536 0.0017 3.162 0.0032 2.887 0.0053 2.673 0.0079 2.500 0.0110 2.357 0.0145 2.236 0.0185
5.500 0.0001 4.491 0.0003 3.889 0.0008 3.479 0.0016 3.175 0.0028 2.940 0.0043 2.750 0.0063 2.593 0.0086 2.460 0.0113
6.000 0.0000 4.899 0.0002 4.243 0.0004 3.795 0.0008 3.464 0.0014 3.207 0.0023 3.000 0.0035 2.828 0.0050 2.683 0.0067
5.307 0.0001 4.596 0.0002 4.111 0.0004 3.753 0.0007 3.474 0.0012 3.250 0.0019 3.064 0.0028 2.907 0.0039
5.716 0.0000 4.950 0.0001 4.427 0.0002 4.042 0.0004 3.742 0.0006 3.500 0.0010 3.300 0.0016 3.131 0.0023
5.657 0.0000 4.743 0.0001 4.330 0.0002 4.009 0.0003 3.750 0.0006 3.536 0.0009 3.354 0.0013
5.060 0.0000 4.619 0.0001 4.276 0.0002 4.000 0.0003 3.771 0.0005 3.578 0.0007
4.908 0.0000 4.543 0.0001 4.250 0.0001 4.007 0.0002 3.801 0.0004
4.811 0.0000 4.500 0.0000 4.243 0.0001 4.025 0.0002
4.750 0.0000 4.478 0.0000 4.249 0.0001
4.714 0.0000 4.472 0.0000
4.696 0.0000

N0 = 10 N0 = 11 N0 = 12 N0 = 13 N0 = 14 N0 = 15 N0 = 16 N0 = 17 N0 = 18 N0 = 19
k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α

0.000 0.5444 0.000 0.5423 0.000 0.5405 0.000 0.5389 0.000 0.5375 0.000 0.5363 0.000 0.5351 0.000 0.5341 0.000 0.5331 0.000 0.5323
0.213 0.4557 0.204 0.4577 0.196 0.4595 0.189 0.4611 0.183 0.4625 0.177 0.4637 0.172 0.4649 0.167 0.4659 0.162 0.4669 0.158 0.4677
0.426 0.3693 0.408 0.3751 0.392 0.3802 0.378 0.3847 0.365 0.3888 0.354 0.3924 0.343 0.3958 0.333 0.3988 0.324 0.4016 0.316 0.4041
0.640 0.2897 0.612 0.2984 0.588 0.3061 0.567 0.3130 0.548 0.3192 0.530 0.3248 0.515 0.3299 0.500 0.3347 0.487 0.3390 0.474 0.3430
0.853 0.2199 0.817 0.2303 0.785 0.2396 0.756 0.2481 0.730 0.2558 0.707 0.2628 0.686 0.2693 0.667 0.2753 0.649 0.2808 0.633 0.2860
1.066 0.1617 1.021 0.1725 0.981 0.1825 0.945 0.1916 0.913 0.2000 0.884 0.2078 0.858 0.2151 0.833 0.2219 0.811 0.2282 0.791 0.2341
1.279 0.1154 1.225 0.1256 1.177 0.1353 1.134 0.1443 1.096 0.1527 1.061 0.1607 1.029 0.1682 1.000 0.1752 0.973 0.1819 0.949 0.1881
1.492 0.0799 1.429 0.0890 1.373 0.0977 1.323 0.1060 1.278 0.1139 1.237 0.1215 1.201 0.1287 1.167 0.1356 1.136 0.1422 1.107 0.1484
1.706 0.0539 1.633 0.0614 1.569 0.0688 1.512 0.0760 1.461 0.0831 1.414 0.0899 1.372 0.0965 1.333 0.1029 1.298 0.1091 1.265 0.1150
1.919 0.0354 1.837 0.0413 1.765 0.0473 1.701 0.0533 1.643 0.0592 1.591 0.0651 1.544 0.0709 1.500 0.0766 1.460 0.0821 1.423 0.0875
2.132 0.0227 2.041 0.0271 1.961 0.0318 1.890 0.0365 1.826 0.0413 1.768 0.0462 1.715 0.0510 1.667 0.0559 1.622 0.0607 1.581 0.0654
2.345 0.0142 2.245 0.0175 2.157 0.0209 2.079 0.0245 2.008 0.0283 1.945 0.0321 1.887 0.0361 1.833 0.0400 1.784 0.0441 1.739 0.0481
2.558 0.0087 2.450 0.0110 2.353 0.0135 2.268 0.0161 2.191 0.0190 2.121 0.0219 2.058 0.0250 2.000 0.0282 1.947 0.0314 1.897 0.0347
2.772 0.0053 2.654 0.0068 2.550 0.0085 2.457 0.0104 2.374 0.0125 2.298 0.0147 2.230 0.0171 2.167 0.0195 2.109 0.0221 2.056 0.0247
2.985 0.0031 2.858 0.0041 2.746 0.0053 2.646 0.0066 2.556 0.0081 2.475 0.0097 2.401 0.0114 2.333 0.0133 2.271 0.0152 2.214 0.0173
3.198 0.0018 3.062 0.0025 2.942 0.0032 2.835 0.0041 2.739 0.0051 2.652 0.0063 2.573 0.0075 2.500 0.0089 2.433 0.0103 2.372 0.0119
3.411 0.0010 3.266 0.0014 3.138 0.0019 3.024 0.0025 2.921 0.0032 2.828 0.0040 2.744 0.0049 2.667 0.0059 2.596 0.0069 2.530 0.0081
3.624 0.0006 3.470 0.0008 3.334 0.0012 3.213 0.0015 3.104 0.0020 3.005 0.0025 2.916 0.0031 2.833 0.0038 2.758 0.0046 2.688 0.0054
3.838 0.0003 3.674 0.0005 3.530 0.0007 3.402 0.0009 3.286 0.0012 3.182 0.0016 3.087 0.0020 3.000 0.0024 2.920 0.0030 2.846 0.0036
4.051 0.0002 3.878 0.0003 3.726 0.0004 3.591 0.0005 3.469 0.0007 3.359 0.0010 3.259 0.0012 3.167 0.0015 3.082 0.0019 3.004 0.0023
4.264 0.0001 4.083 0.0001 3.922 0.0002 3.780 0.0003 3.652 0.0004 3.536 0.0006 3.430 0.0008 3.333 0.0010 3.244 0.0012 3.162 0.0015
4.477 0.0000 4.287 0.0000 4.118 0.0001 3.969 0.0002 3.834 0.0003 3.712 0.0003 3.602 0.0005 3.500 0.0006 3.407 0.0008 3.320 0.0009
4.690 0.0000 4.491 0.0000 4.315 0.0000 4.158 0.0001 4.017 0.0001 3.889 0.0002 3.773 0.0003 3.667 0.0004 3.569 0.0005 3.479 0.0006
4.511 0.0000 4.347 0.0000 4.199 0.0000 4.066 0.0001 3.945 0.0001 3.833 0.0002 3.731 0.0003 3.637 0.0004
4.536 0.0000 4.382 0.0000 4.243 0.0000 4.116 0.0000 4.000 0.0001 3.893 0.0002 3.795 0.0002
4.419 0.0000 4.288 0.0000 4.167 0.0000 4.056 0.0001 3.953 0.0001
4.459 0.0000 4.333 0.0000 4.218 0.0000 4.111 0.0000
4.380 0.0000 4.269 0.0000

N0 = 20 N0 = 21 N0 = 22 N0 = 23 N0 = 24 N0 = 25 N0 = 26 N0 = 27 N0 = 28 N0 = 29 N0 = 30
k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α

0.000 0.5314 0.000 0.5307 0.000 0.5300 0.000 0.5293 0.000 0.5287 0.000 0.5281 0.000 0.5276 0.000 0.5271 0.000 0.5266 0.000 0.5261 0.000 0.5257
0.154 0.4686 0.151 0.4693 0.147 0.4700 0.144 0.4707 0.141 0.4713 0.139 0.4719 0.136 0.4724 0.134 0.4729 0.131 0.4734 0.129 0.4739 0.127 0.4743
0.309 0.4065 0.302 0.4087 0.295 0.4107 0.289 0.4127 0.283 0.4145 0.277 0.4162 0.272 0.4178 0.267 0.4193 0.263 0.4207 0.258 0.4221 0.254 0.4233
0.463 0.3468 0.452 0.3503 0.442 0.3535 0.433 0.3566 0.424 0.3594 0.416 0.3621 0.408 0.3647 0.401 0.3671 0.394 0.3694 0.387 0.3716 0.381 0.3737
0.617 0.2908 0.603 0.2953 0.590 0.2996 0.577 0.3035 0.566 0.3073 0.555 0.3108 0.544 0.3142 0.535 0.3174 0.525 0.3204 0.516 0.3233 0.508 0.3261
0.772 0.2397 0.754 0.2449 0.737 0.2498 0.722 0.2545 0.707 0.2589 0.693 0.2631 0.680 0.2671 0.668 0.2709 0.657 0.2745 0.646 0.2779 0.635 0.2812
0.926 0.1941 0.905 0.1997 0.885 0.2050 0.866 0.2101 0.849 0.2149 0.832 0.2195 0.817 0.2239 0.802 0.2281 0.788 0.2321 0.775 0.2360 0.762 0.2397
1.080 0.1544 1.055 0.1601 1.032 0.1656 1.010 0.1708 0.990 0.1758 0.971 0.1806 0.953 0.1852 0.935 0.1896 0.919 0.1938 0.904 0.1979 0.889 0.2018
1.234 0.1207 1.206 0.1262 1.180 0.1316 1.155 0.1367 1.131 0.1416 1.109 0.1464 1.089 0.1510 1.069 0.1555 1.051 0.1597 1.033 0.1639 1.016 0.1679
1.389 0.0928 1.357 0.0979 1.327 0.1029 1.299 0.1077 1.273 0.1124 1.248 0.1170 1.225 0.1215 1.203 0.1258 1.182 0.1300 1.162 0.1340 1.143 0.1380
1.543 0.0701 1.508 0.0747 1.474 0.0792 1.443 0.0836 1.414 0.0879 1.387 0.0922 1.361 0.0963 1.336 0.1004 1.313 0.1044 1.291 0.1082 1.270 0.1120
1.697 0.0521 1.658 0.0561 1.622 0.0600 1.588 0.0639 1.556 0.0678 1.525 0.0716 1.497 0.0754 1.470 0.0791 1.444 0.0827 1.420 0.0863 1.397 0.0898
1.852 0.0381 1.809 0.0414 1.769 0.0448 1.732 0.0481 1.697 0.0515 1.664 0.0549 1.633 0.0582 1.604 0.0615 1.576 0.0647 1.549 0.0680 1.524 0.0712
2.006 0.0274 1.960 0.0301 1.917 0.0329 1.876 0.0357 1.839 0.0386 1.803 0.0414 1.769 0.0443 1.737 0.0472 1.707 0.0500 1.678 0.0529 1.651 0.0557
492 A. Vivier and J. Aupiais

N0 = 20 N0 = 21 N0 = 22 N0 = 23 N0 = 24 N0 = 25 N0 = 26 N0 = 27 N0 = 28 N0 = 29 N0 = 30
k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α k1−α α

2.160 0.0194 2.111 0.0216 2.064 0.0238 2.021 0.0261 1.980 0.0285 1.942 0.0309 1.905 0.0333 1.871 0.0357 1.838 0.0382 1.807 0.0406 1.778 0.0431
2.315 0.0135 2.261 0.0152 2.212 0.0170 2.165 0.0189 2.121 0.0208 2.080 0.0227 2.041 0.0247 2.005 0.0267 1.970 0.0288 1.937 0.0309 1.905 0.0330
2.469 0.0093 2.412 0.0106 2.359 0.0120 2.309 0.0134 2.263 0.0149 2.219 0.0165 2.177 0.0181 2.138 0.0197 2.101 0.0214 2.066 0.0231 2.032 0.0249
2.623 0.0063 2.563 0.0073 2.507 0.0083 2.454 0.0094 2.404 0.0106 2.358 0.0118 2.313 0.0131 2.272 0.0144 2.232 0.0158 2.195 0.0172 2.159 0.0186
2.778 0.0042 2.714 0.0049 2.654 0.0057 2.598 0.0065 2.546 0.0074 2.496 0.0084 2.450 0.0094 2.405 0.0104 2.364 0.0115 2.324 0.0126 2.286 0.0137
2.932 0.0028 2.864 0.0033 2.801 0.0039 2.742 0.0045 2.687 0.0051 2.635 0.0059 2.586 0.0066 2.539 0.0074 2.495 0.0082 2.453 0.0091 2.413 0.0100
3.086 0.0018 3.015 0.0022 2.949 0.0026 2.887 0.0030 2.828 0.0035 2.774 0.0040 2.722 0.0046 2.673 0.0052 2.626 0.0059 2.582 0.0065 2.540 0.0073
3.240 0.0012 3.166 0.0014 3.096 0.0017 3.031 0.0020 2.970 0.0024 2.912 0.0028 2.858 0.0032 2.806 0.0036 2.757 0.0041 2.711 0.0046 2.667 0.0052
3.395 0.0007 3.317 0.0009 3.244 0.0011 3.175 0.0013 3.111 0.0016 3.051 0.0019 2.994 0.0022 2.940 0.0025 2.889 0.0029 2.840 0.0033 2.794 0.0037
3.549 0.0005 3.467 0.0006 3.391 0.0007 3.320 0.0009 3.253 0.0011 3.190 0.0013 3.130 0.0015 3.074 0.0017 3.020 0.0020 2.969 0.0023 2.921 0.0026
3.703 0.0003 3.618 0.0004 3.539 0.0005 3.464 0.0006 3.394 0.0007 3.328 0.0008 3.266 0.0010 3.207 0.0012 3.151 0.0013 3.098 0.0016 3.048 0.0018
3.858 0.0002 3.769 0.0002 3.686 0.0003 3.608 0.0004 3.536 0.0004 3.467 0.0005 3.402 0.0006 3.341 0.0008 3.283 0.0009 3.228 0.0011 3.175 0.0012
4.012 0.0001 3.920 0.0001 3.834 0.0002 3.753 0.0002 3.677 0.0003 3.606 0.0003 3.538 0.0004 3.474 0.0005 3.414 0.0006 3.357 0.0007 3.302 0.0008
4.166 0.0000 4.070 0.0000 3.981 0.0001 3.897 0.0001 3.818 0.0002 3.744 0.0002 3.674 0.0003 3.608 0.0003 3.545 0.0004 3.486 0.0005 3.429 0.0006
4.321 0.0000 4.221 0.0000 4.128 0.0000 4.042 0.0000 3.960 0.0000 3.883 0.0001 3.810 0.0002 3.742 0.0002 3.677 0.0003 3.615 0.0003 3.556 0.0004
4.372 0.0000 4.276 0.0000 4.186 0.0000 4.101 0.0000 4.022 0.0000 3.946 0.0001 3.875 0.0001 3.808 0.0002 3.744 0.0002 3.683 0.0002
4.330 0.0000 4.243 0.0000 4.160 0.0000 4.083 0.0000 4.009 0.0000 3.939 0.0001 3.873 0.0001 3.810 0.0001
4.299 0.0000 4.219 0.0000 4.143 0.0000 4.071 0.0000 4.002 0.0000 3.937 0.0001
4.276 0.0000 4.202 0.0000 4.131 0.0000 4.064 0.0000
4.333 0.0000 4.260 0.0000 4.191 0.0000
4.389 0.0000 4.318 0.0000
4.445 0.0000

References application to unfolding of spectrometric measurements without

the influence of sample treatment. Genève (2005).
1. Neuilly, M.: Statistique Appliquée à l’Exploitation des Mesures. 7. Knoll, G. F.: Radiation Detection and Measurement. John Wiley
Masson, Paris (1986). & Sons, New York (1979).
2. Michel, R.: Quality assurance of nuclear analytical techniques 8. Saporta, G.: Probabilités Analyse des Données et Statistique. Paris
based on Bayesian characteristic limits. J. Radioanal. Nucl. Chem. (1990).
245, 137 (2000). 9. Rainwater, L. J., Wu, C. S.: Applications of probability theory to
3. SO: Determination of the detection limit and decision threshold nuclear particle detection. Nucleonics 1, 60 (1947).
for ionizing radiation measurements – Part 7: Fundamentals and 10. Little, R. J. A.: The statistical analysis of low-level radioactivity
general applications. Genève (2005). in the presence of background counts. Health Phys. 43, 693
4. ISO: Determination of the detection limit and decision threshold (1982).
for ionizing radiation measurements – Part 5: Fundamentals and 11. Weise, K.: Bayesian-statistical decision threshold, detection limit,
applications to counting measurements on filters during accumu- and confidence interval in nuclear radiation measurement. Kern-
lation of radioactive material. Genève (2005). technik 63, 214 (1998).
5. ISO: Determination of the detection limit and decision threshold 12. Brüchle, W.: Confidence intervals for experiments with back-
for ionizing radiation measurements – Part 6: Fundamentals and ground and small numbers of events. Radiochim. Acta 91, 71
applications to measurements by use of transient mode. Genève (2003).
(2005). 13. Miller, G., Martz, H. F., Little, T. T., Guilmette, R.: Using exact
6. SO: Determination of the detection limit and decision threshold poisson likelihood functions in bayesian interpretation of counting
for ionizing radiation measurements – Part 8: Fundamentals and measurements. Health Phys. 83, 512 (2002).