Scribd
Upload
174 views

Lecture 4 (Statistical Testing)

The document discusses statistical testing procedures for hypotheses about population parameters. It describes: 1) How statistical testing involves making a statement (null hypothesis) about a population parameter, collecting a sample, computing a test statistic based on the sample, and either accepting or rejecting the null hypothesis. 2) Common outcomes of hypothesis testing include correctly accepting or rejecting the null hypothesis, or making type I or type II errors. The significance level determines the probability of a type I error. 3) Procedures for testing hypotheses about the mean using the sample mean as the test statistic, and about the variance using the sample variance as the test statistic. Critical values define the acceptance regions for these tests.

Uploaded by

Kismet
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
174 views

Lecture 4 (Statistical Testing)

The document discusses statistical testing procedures for hypotheses about population parameters. It describes: 1) How statistical testing involves making a statement (null hypothesis) about a population parameter, collecting a sample, computing a test statistic based on the sample, and either accepting or rejecting the null hypothesis. 2) Common outcomes of hypothesis testing include correctly accepting or rejecting the null hypothesis, or making type I or type II errors. The significance level determines the probability of a type I error. 3) Procedures for testing hypotheses about the mean using the sample mean as the test statistic, and about the variance using the sample variance as the test statistic. Critical values define the acceptance regions for these tests.

Uploaded by

Kismet
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Purpose of Statistical Testing

• To determine whether or not a


population has a particular probability
distribution using a sample
• To determine whether a parameter has
a particular value based on a sample

Statistical Testing
(For the Mean and Variance)

Statistical Testing Fallibility of the Null Hypothesis


1. Make a statement (statistical • The testing of a null hypothesis is not
hypothesis/null hypothesis, H0) about infallible since it is based on a sample
the probability distribution drawn from the population, and not on
2. Test the hypothesis on a sample of the the entire population
population by computing a sample
statistic (test statistic)
3. Accept or reject the hypothesis based
on the value of the test statistic
4. If the null hypothesis is rejected, the
complementary hypothesis
(alternative hypothesis, H1) will be
accepted

Possible Outcomes for H0 Size of Type I Error


1. H0 is accepted, when H0 is TRUE • probability of rejecting H0 when H0 is
2. H0 is rejected, when H0 is TRUE true
3. H0 is accepted, when H0 is FALSE
α = P[Reject H0 when H0 is true]
4. H0 is rejected, when H0 is FALSE
Significance level of the test – when α is
Outcomes 1 and 4: no error is made; fixed at some level for H0 and is
correct course of action expressed as some percentage
Outcome 2: Type I error – equivalent to (1 – confidence level)
Outcome 3: Type II error – commonly used values are 5%
(significant) and 1% (highly significant)

1
Test of the Mean of a Probability Test of the Mean of a Probability
Distribution Distribution
• The hypothesis that the mean (µ) of a • Under the null hypothesis (µ = µ0), the
probability distribution is equal to a probability that the sample mean (and
specific value (µ0) can be tested by therefore the population mean) lies
using the sample mean X as the test within the acceptable range is:
statistic.
H0: µ = µ0 P[( µ0 − c) < X < ( µ0 + c)] = 2φ ( z ) − 1
H1: µ ≠ µ0
where: zσ
** X is assumed to be normally c=
n
distributed, or at least approximately
normally distributed

Test of the Mean of a Probability Test of the Mean of a Probability


Distribution Distribution
• If σ is unknown, the probability • H0 is accepted if X lies between (µ0 – c)
statement becomes: and (µ0 + c)
• Otherwise, H0 is rejected
P[( µ0 − c) < X < ( µ0 + c)] = 2 F (t ) − 1

where: ts
c=
n

Test of the Mean of a Probability Test of the Mean of a Probability


Distribution Distribution
f (X ) • If α is the probability that H0 is rejected
when it is true, then 1-α would be the
Reject H0 Accept H0 Reject H0
probability that H0 is accepted when it
is true
• Therefore,
1-α = 2Φ(z)-1 (for known σ)
= 2F(t)-1 (for unknown σ)
α α
2 2

µ0 - c µ0 µ0 + c X

2
Test of the Mean of a Probability
Distribution
Example
• Thus, we can solve for Φ(z) and F(t): An angle is measured 10 times. Each
α measurement is independent and made
φ (z) = 1 − (for known σ) with the same precision, i.e., the 10
2
measurements constitute a random
α
F (t ) = 1 − (for unknown σ) sample of size 10. The sample mean
2 and sample standard deviation are
• After we get the value of z or t, which is calculated from the measurements as:
Φ(z) or F(t), we can then derive z or t x=42°12’16” and S=3.7”. Test at the
from the table of values 5% significance level the hypothesis
that the population mean µ is
42°12’16.0”, against the alternative
that µ is not equal to 42°12’16.0”.

Solution Solution
For 5% level of significance, α = 0.05 The limits (critical values) can then be computed
α as:
p = F (t ) = 1 − = 0.975 ( µ 0 − c) = 42°12 '13.4"
2
degrees of freedom = n − 1 = 10 − 1 = 9 ( µ 0 + c) = 42°12 '18.6"

From the table, t = t0.975,9 = 2.26


Given that the sample mean is 42°12’14.6”
Therefore, and this value lies within the range of the
limits (critical region), then the null
ts (2.26)(3.7") hypothesis that the population mean (µ) =
c= = = 2.6" 42°12’14.6” is accepted at 5% level of
n 10 significance.

Test of the Variance of a Test of the Variance of a


Probability Distribution Probability Distribution
• Under the assumption that a population • Under the null hypothesis (σ2 = σ02),
is normally distributed, the null the probability that the sample variance
hypothesis (H0) that the population (and therefore the population variance)
variance σ2 is equal to σ02 can be tested lies within the acceptable range is:
against the alternative hypothesis (that
σ2 is not equal to σ02) using the sample χ2 σ 2 χ2 σ 2 
P  a ,n −1 0 < S 2 < b,n −1 0  = b − a
variance S2 as the test statistic  ( n − 1) (n − 1) 

3
Test of the Variance of a Test of the Variance of a
Probability Distribution Probability Distribution
• H0 is accepted if S2 lies between χ a ,n −1σ 0 f (s 2 )
2 2

χ 2
σ
and b ,n −1 0
2
( n − 1)
Reject H0 Accept H0 Reject H0
(n − 1)

• Otherwise, H0 is rejected

α α
2 2

 χ12−(α / 2), n-1  2


  σ 0
σ 02  χ (2α / 2), n-1  2
σ
s2
 n −1 
 n − 1  0
 

Test of the Variance of a


Probability Distribution
Example
• Again, if 1-α is the probability that H0 is An angle is measured 10 times. Each
accepted when it is true, therefore: measurement is independent and made
1-α = b-a with the same precision, i.e., the 10
measurements constitute a random
sample of size 10. The sample standard
• Given that a+b=1 (complementary deviation is calculated from the
percentiles), we can solve for the measurements as S=3.7”. Test at the
percentiles: 5% significance level the hypothesis
α α that the population standard deviation
a= b = 1− is 2.0”, against the alternative that σ is
2 2 not equal to 2.0”.

Solution Solution
For 5% level of significance, α = 0.05 The limits (critical values) are therefore,
α α χ 0.025,9
2
σ 02 (2.70)(2.0") 2
a= = 0.025 and b = 1− = 0.975 = = 1.2 (seconds of arc) 2
2 2 n −1 9
degrees of freedom = n − 1 = 10 − 1 = 9
χ 0.975,9
2
σ 02 (19.0)(2.0") 2
= = 8.44 (seconds of arc) 2
From the table, χ20.025,9 = 2.70 and χ20.975,9 = n −1 9
19.0 Given that the sample variance S2 =
(3.7)2 = 13.7 (seconds of arc)2 and this
value does not lie within the range of
the limits (critical region), then the null
hypothesis that the population
standard deviation (σ) = 2.0” is
rejected at 5% level of significance.

You might also like