Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Error Analysis: an introduction

”To err is human; to describe the error properly is
sublime”.....Cliff Swartz.

Department of Physics

BITS-Pilani, K.K.Birla Goa campus

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Physics Laboratory 1

Data Analysis: Error estimation

The error in an experimentally measured quantity is NEVER found by comparing it
with a number measured by somebody else!!.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Outline

1 Motivation

2 Experimental Errors: Systematic Error

3 Experimental Errors: Random Error

4 Accuracy, Precision, Significant figures and Round-off

5 Propagation of Errors of Precision

6 Linear fit to a set of data points

7 Data analysis using openoffice

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Motivation

Error, the difference between a calculated or observed value

and the true value of a quantity, is INEVITABLE!
The purpose of this presentation is
to familiarize you with the types of experimental errors that you will
encounter in all the experiments to be done in this lab and
to show you how to calculate the errors in the measured experimental
data.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Experimental Errors: Systematic Error

Almost all experimental uncertainties are due to the presence of two types of Errors:
Systematic errors and Random errors.
1 Systematic Errors :

They are mostly due to defects in the measuring devices.

Instrument calibration: The needle of a meter does not point to the 0 of the scale (zero error), the
clock might run faster always etc .
External factors might influence the experiment and lead to inaccuracies.
In an experiment to measure the magnetic field of a bar magnet using a magnetometer, the bar
magnet is to be oriented perpendicular to the earth’s magnetic field in order to cancel the effect of
the earth’s magnetic field .
They are inaccuracies causing the measurements to shift constantly in the same
direction (either too high or too low).

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Experimental Errors: Random Error

1 Random Error :

Simple Pendulum experiment: Time T5 (in sec) of 5 complete oscillations is

recorded 11 times. The results are tabulated below in Table 1. (We rule out any
possibility of error due to negligence).
Table 1
Sr.No. 1 2 3 4 5 6 7 8 9 10 11
T5 (s) 4.30 4.28 4.32 4.29 4.31 4.29 4.30 4.31 4.30 4.33 4.27

Contrary to our expectations, the measured T5 values are not identical.

Each of the measured T5 value seems random and is subject to a random error.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Normal distribution: Random error

This data in the table is displayed in the figure below:

Figure: The frequency (no. of times that a certain T5 is repeated) is plotted as a function of T5 (s).

From the histogram (on the top of which a best fit curve is shown), we see that
T5 has a peak around 4.30 s. It is the most probable(mean) value that
appears with highest frequency (probability).
On both sides of the mean/peak, the probability decreases.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Mean and Standard Division

When a measurement is repeated a large number of times, we see that the measured
values are grouped around some central value. This grouping or distribution can be
described with two numbers:
The mean (µ), which measures the central value.
The standard deviation (σ) which describes the spread or deviation of the
measured values from the mean.
Mean:
For a set of N measured values {xi , i = 1, .... N} of a quantity x, the mean µ is
calculated as:
N
1 X 1
µ = xbest = xi = (x1 + x2 + .... + xN−1 + xN )
N i=1 N

Standard deviation(SD):
The standard deviation σ of the measured values is calculated as
v
u
u 1 X N
σ=t (xi − µ)2 .
N − 1 i=1

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Mean, SD and true value

The error or uncertainty in the mean value is the standard error (random), σm ,
defined as: v
u
u1 X M  2
(i)
δx = σm (= σSDOM ) = t xbest − µ ,
N i=1

where M
1 X (i)
µ= x ,
M i=1 best

M being the number of identical set of observations. It can be shown that

σ
σm = √ .
N

Finally for a single set of measurement, you quote

√
xreport = xbest + δx = µ ± σ/ N

Note that in the limit N → ∞, xbest → xtrue and δx → 0.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Normal distribution....continued

The Probability distribution of quantities(time, length,current,...etc) subject to

random errors is characterized by the mean and the standard deviation. Such a
distribution is known as Gaussian distribution(Normal distribution).
For measurements (of N → ∞) of any such physical quantity x, the probability
2 2
of obtaining a value xi is given by, P(xi ) = √1 e (xi −µ) /(2σ )
σ 2π
For a single measurement, the probability
P(xi ) for finding a single measurement xi Normal Distribution
to be within the range (µ − tσ, µ + tσ) is
equal to 68.2%, 95.4% and 99.7% for
t = 1, 2 and 3 respectively.

¿¿
Department of Physics BITS-Pilani, K.K.Birla Goa campus
Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Significant figure, Accuracy, Precision, Round-off

1 Accuracy: Accuracy measures how close a measured value is to the true value.
2 Precision: Precision(“reproducibility”) measures how closely two or more
measurements agree with each other.
If the g value is 9.80 ± 0.01 m/s 2 and g = 9.789 ± 0.001 m/s2 , the first
one is more accurate, whereas the second one is more precise.
3 Significant figures/Round off: The number of significant figures in a measured
number is equal to number of digits that are known with some degree of
reliability.
Say, Munna reported g = 9.82 ± 0.02385 m/s2 .
Absurd! If g is doubtful from the third decimal, how can we know the
uncertainty all the way to the fifth decimal place?

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Significant figures/Round off ...continued

1 The result should be stated as g = 9.82 ± 0.02 m/sec2 . That is, the error
should be rounded off at the 2nd decimal.
2 The number of significant figures(both in mean and error) is 3.
3 Note the following

(36.479 × 2.6)/14.85 = 6.3868956 = 6.4 (approximately), because the

original factor 2.6 has only two significant digits.
16.345 + 1.4 − 5.23 = 12.515 = approximately 12.5, because the original
term has only one significant digit after the decimal point.
Error should always be reported to one or two significant
figures.
Problem 1: Suppose, Munna reported the speed(u) of a object as u = 6051.78 ± 30 m/sec. Is it correct?
Problem 2: Express the following to the correct number of significant figures:
a. 25.052 ± 1.502
b. 92 ± 3.14159

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Propagation of Errors

The accuracy of a result is limited by the propagated error.

1 Suppose q = x + y and we measure x and y as xbest + δx and ybest + δy ,
respectively.
Q. What is qbest and the error δq?
A. We find
q = (xbest + δx) + (ybest + +δy ) = qbest + δq

where qbest = xbest + ybest and δq ' δx + δy

2 If both x and y are independent and errors associated with them are random,
then δq 6= δx + δy .
3 They are governed by the Gaussian distribution and one finds

p
δq = (δx)2 + (δy )2 → i.e. errors are added in quadrature.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Propagation of Errors

1 Suppose q = x.y and we measure x and y as xbest + δx and ybest + δy ,

respectively.
Q. What is qbest and the error δq?
A. We find

q = (xbest + δx).(ybest + +δy ) = qbest + δq

where qbest = xbest ybest and

δq δx δy
δq ' (xbest δy + ybest δx) → qbest
' xbest
+ ybest

If x and y obeys Gaussian distribution, one finds

s 2  2
δq δx δy
' +
qbest xbest ybest

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Linear fit to a set of data points

1 Consider a set of experimental data (x1 , y1 ), (x2 , y2 ), .....(xN , yN ) where x and

y measurements have random errors.
2 We expect to have a linear relationship between x and y of the form
y = mx + c where c and m are unknown constants, known to be the
intercept and the slope of the best fit line.
3 Slope(m) and intercept(c) of the best fit line:
Let Si = yi − mxi − c be the deviation of any experimental point (xi , yi )
from the best fit line and define
XN
S= (yi − mxi − c)2
i=1

For best fits of our data S is minimum i.e.

∂S X ∂S X
= −2 xi (yi − mxi − c) = 0 and = −2 (yi − mxi − c) = 0.
∂m ∂c

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Linear fit to a set of data points

1 Slope(m) and intercept(c) of the best fit line:

The second equation can be written as y = mx + c where y = N1
P
yi
1 P
and x = N xi i.e. the best fit line passes through the centroid (x, y ) of
the points (xi , yi )
Solving the above two equations, we find
P
P P
xi yi −
N xi yi
m=
∆
P 2
xi − ( xi )2 .
P
where c = y − mx and ∆ = N
The line that best fits the data is given by the m and c values given above.

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Data analysis using openoffice

Important things to learn and master (To know more about openoffice,
see http://www.tutorialsforopenoffice.org).

1 Every single cell can be used as a calculator

2 Copying and/or ”incrementing” cell contents along rows or columns
3 Creating scatter plots
4 Mathematical and statistical functions.
In particular, the LINEST function.
1 =average(data-range) −→ finds the average of the data
2 =stdev(data-range) −→ finds the std dev of the data
3 =index(linest(y-range,x-range,1,1),1,1) −→ finds the best fit slope
4 =index(linest(y-range,x-range,1,1),2,1) −→ finds the error in the best fit slope
5 =index(linest(-range,x-range,1,1),1,2) −→ finds the best fit intercept
6 =index(linest(y-range,x-range,1,1),2,2) −→ finds the error in the best fit
intercept

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction
Outline Motivation Experimental Errors: Systematic Error Experimental Errors: Random Error Accuracy, Precision, Significant fig

Thank You for your kind attention!

Enjoy!

Department of Physics BITS-Pilani, K.K.Birla Goa campus

Error Analysis: an introduction