1

Data Handling and Measurement

A. Errors and Uncertainty
With every measurement, no matter how carefully it is made, there is an associated error and
uncertainty inherent with the measurement. No one can ever exactly measure the true value
of a quantity. An experimental error just means that there is a difference between the
recorded value and the perfect or correct value.
The magnitude of the error is due to the precision of the measuring device, the proper
calibration of the device, and the competent application of the device. Understanding possible
errors is an important issue in any experimental science. The conclusions we draw from the
data, and especially the strength of those conclusions, will depend on how well we control the
uncertainties.
Generally, error and uncertainty are two different terms. Error is the difference between
the measured value and the true value of the thing being measured. On the other hand,
uncertainty is a quantification of the doubt about the measurement result. Whenever possible
we try to correct for any known errors: for example, by calibrating the instruments. But any
error whose value we do not know is a source of uncertainty.

1. Random and Systematic Errors
There are two types of measurement errors random and systematic.
a. Random Error
If you measure a quantity many times and get lots of slightly different readings then
this called a random error. For example, when measuring the bounce of a ball it is
very difficult to get the same value every time even if the ball is doing the same thing.
Sources of random errors include:
i. The readability of the instrument
ii. The observer being less than perfect
iii. The effects of change in the surroundings
b. Systematic Error
This is when there is something wrong with the measuring device or method. Using a
ruler with a broken end can lead to a zero error. Even with no random error in the
results, youd still get the wrong answer.


2

Sources of systematic errors include:
i. An instrument with zero error. To correct for zero error, the value should be
subtracted from every reading
ii. An instrument being wrongly calibrated
iii. The observer being less than perfect in the same way every measurement
c. Reducing Errors
To reduce random errors you can repeat your measurements. If the uncertainty is truly
random, they will lay either side of the true reading and the mean of these values will
be close to the actual value.
To reduce a systematic error you need to find out what is causing it and correct your
measurements accordingly. A systematic error is not easy to spot by looking at the
measurements, but is sometimes apparent when you look at the graph of your results
or the final calculated value.

2. Accuracy and Precision*
*Note : Our SPM Curriculum Specification uses the term consistency instead of
precision
To determine the error associated with a measurement, scientists often refer to two terms that
are different from each other, the precision and accuracy of the measurement. To better
understand the difference between these two properties, let us use the analogy of the
marksman who uses a gun to fire bullets at a target. In this analogy, the gun is the instrument,
the marksman is the operator of the instrument, and the results are determined by the location
of the bullet holes in the target.
a. Precision
The precision of an experiment is a measure of the reliability of the experiment, or
how reproducible the experiment is. In Figure 1, we see that the marksmans
instrument was quite precise, since his results were uniform due to the use of a
sighting scope. However, the instrument did not provide accurate results since the
shots were not centered on the targets bulls eye. The fact that his results were
precise, but not accurate, could be due to a misaligned sighting scope, or a consistent
operator error. Therefore precision tells us something about the quality of the
instruments operation.
b. Accuracy
The accuracy of an experiment is a measure of how closely the experimental results
agree with a true or accepted value. In Figure 2, we see a different experimental

3

result. Here, the shots are centred on the bulls eye but the results were not uniform,
indicating that the marksmans instrument displayed good accuracy but poor
precision. This could be the result of a poorly manufactured gun barrel. In this case,
the marksman will never achieve both accuracy and precision, even if he uses the
instrument very carefully. If he is not satisfied with the results he must change his
equipment. Therefore accuracy tells us something about the quality or correctness
of the result.

c. Errors - Accuracy and Precision Relation
As scientists, we desire our results to be both precise and accurate. One benefit of
taking many measurements of a single property is that human or operator errors are
easily detected. Because several measurements were made, we can discount the errant
data point as an obvious mistake that is probably due to operator error. But,
systematic errors cannot be reduced by repeating the experiment.
An accurate experiment is one that has a small systematic error, whereas a precise
experiment is one that has a small random error.

3. Estimating The Uncertainty Range
An uncertainty range applies to any experimental value. The idea is that, instead of just
giving one value that implies perfection, we give the likely range of measurement.
a. Significant Digits
Any experimental measurement should be quoted with its uncertainty. This indicates
the possible range of values for the quantity being measured. At the same time, the
number of significant digits used will act as a guide to the amount of uncertainty. For
example, a measurement of mass which is quoted as 23.456 g implies an uncertainty
of 0.001 g (it has five significant digits), whereas one of 23.5 g implies an
uncertainty of 0.1 g (it has three significant digits)

Figure 1: High
precision and low
accuracy
Figure 2: High
accuracy and low
precision
Figure 3: High
precision and accuracy.
Figure 4: High
precision and accuracy
with an operator error.

4

A simple rule of calculations (multiplication or division) is to quote the answer to the
same number of significant digits as the LEAST precise value that is used.
b. Uncertainties Representation.
Uncertainty can be written in 3 formats. If a value of a measurement is 6.25 m and the
uncertainty is 0.05 cm, this uncertainty can be represented in the forms below :
Absolute uncertainty : 6.25 0.05 cm
Relative uncertainty : 6.25 (1.00 0.008) cm
Fractional (percentage) uncertainty : 6.25 (1.00 0.008) cm = 6.25 0.8%

c. Determining the Absolute Uncertainty from a Single Reading on an
Incremental Scale
When taking a measurement from an instrument (analog or digital) there is a "rule of
thumb", non-statistical method which enables the determination of the absolute
uncertainty without the repetition of measurements. Although the statistical method
referred to above is more reliable than this method, there are times when the rule of
thumb is useful.
First, the term "least count" must be defined. Least count refers to the smallest
increment of measure on the measuring device used. As an example, on a meter stick,
the smallest increment of measure is typically a millimeter. So the least count would
be one millimeter and the absolute uncertainty associated with a meter stick reading
would be one half a millimeters.
Normally the uncertainty range due to readability is estimated as below:
Device Example Uncertainty
Analogue scale Rulers, meters with moving
pointers
half the smallest scale
Digital scale Top-pan balances, digital meters the smallest scale division
Table 1: Uncertainty based on measuring device
d. Estimating Uncertainty Range from Several Repeated Measurements
If the time taken for a trolley to go down a slope is measured five times, the readings
in seconds might be 2.01, 1.82, 1.97, 2.16 and 1.94. The average of these five
readings is 1.98 s. The deviation of the largest and smallest readings can be calculated
(2.16 1.98 = 0.18; 1.98 1.82 = 0.16). The largest value is taken as the uncertainty
range. In this example the time is 1.98 s 0.18s

5

e. Propagation of Uncertainties (how uncertainty in a measurement affects
the final/derived result)
Often, we are not directly interested in a measured value, but we want to use it in a
formula to calculate another quantity. In many cases, we measure many of the
quantities in the formula and each has an associated uncertainty. We deal here with
how to propagate uncertainties to obtain a well-defined uncertainty on a computed
quantity.
Rule 1: Adding/Subtracting Quantities
Whether we add or subtract quantities, the uncertainties must always be added
(never subtracted) to obtain the absolute uncertainty on the computed quantity.
Note that uncertainties on quantities used in a mathematical relationship
always increase the uncertainty on the result. The quantity with the biggest
uncertainty usually dominates the final result. Often one quantity will have a
much bigger uncertainty than all the others. In such cases, we can simply use
this main contribution.
Take as example measuring the length of a car. We measure the distance
between the left wall and the back of the car and subtract the distance from the
wall to the cars front bumper.

So the total length of the car is
( ) ( )
( ) ( )
( ) m 07 . 0 07 . 3
m 02 . 0 05 . 0 m 76 . 2 83 . 5
m 02 . 0 76 . 2 m 05 . 0 83 . 5 length
=
+ =
=


Rule 2: Multiplying/Dividing Quantities
When we multiply or divide quantities, we add (never subtract!) the relative
uncertainties to obtain the relative uncertainty on the computed quantity.
(2.76 0.02) m
(5.83 0.05) m

6

Take as example the area of a rectangle, whose individual sides are measured
to be
a = 25.0 0.5 cm = 25.0 cm (1.00 0.02)
b = 10.0 0.3 cm = 10.0 cm (1.00 0.03).
The area is obtained as follows
( ) ( )
( ) ( )
( ) ( ) ( )
( )
2
2
2
13 250
5 . 12 0 . 250
05 . 0 00 . 1 0 . 250
03 . 0 02 . 0 00 . 1 0 . 10 0 . 25
03 . 0 00 . 1 0 . 10 02 . 0 00 . 1 0 . 25
3 . 0 0 . 10 5 . 0 0 . 25
cm
cm
cm
cm cm
cm cm
cm cm
=
=
=
+ =
=
= Area

Note that the final step has rounded both the result and the uncertainty to an
appropriate number of significant digits.

Rule 3 : For Uncertainty in A Quantity Raised to A Power
Convert absolute uncertainty to fractional (percent) uncertainty and multiply
the fractional uncertainty by the value of the power.
If a = 25.0 0.5 cm, find a
2
.
| |
( ) | |
| |
( )
( )
2
2
2
2
2 2
2
2 2
cm 25 625
cm 04 . 0 00 . 1 0 . 625
cm % 4 0 . 625
cm % 2 2 0 . 625
cm % 2 0 . 25
02 . 0 00 . 1 cm 0 . 25
cm 5 . 0 0 . 25
=
=
=
=
=
=
= a

f. Graphical Representation of Uncertainty
In many situations the best method of presenting and analysing data is to use graph. If
this is the case, a neat way of representing the uncertainties is to use error bars. The
graphs below explain their use.

7

Since the error bar represents the uncertainty range, the best-fit line of the graph
should pass through ALL of the error rectangles created by the error bars (ie. vertical
error bar & horizontal error bar for a graph point).

Figure 5: Error bars (rectangles) representing uncertainties in graph

The best fit line is included by all the error bars in the upper two graphs. This is not true
in the lower graph.
When a quantity is graphed, it is common for the uncertainty of that quantity to be
represented by error bars. Please refer to http://peltiertech.com/WordPress/custom-
error-bars-in-excel-charts/ on how to include error bars into your laboratory reports. You
should know that if a quantity has the same uncertainty value for each measurement, for
example mass, you can enter a fixed error value. If each data point has a unique
uncertainty value, for example volume, you should store those in a column in the
spreadsheet and enter the cells of the custom values.





8

Simplified Notes on Propagating Uncertainties:

1. Adding and subtracting uncertainties:
a. Add absolute uncertainties (in quadrature) to get a single uncertainty.
b. Convert absolute uncertainty to fractional (per cent) uncertainty

2. Multiplying and dividing uncertainties
a. Convert absolute uncertainties to fractional (percent) uncertainties
b. Add fractional uncertainties (in quadrature) to get a single uncertainty.
c. Calculate (absolute) uncertainty of product.

3. Uncertainty in a quantity raised to a power
a. Convert absolute uncertainty to fractional (percent) uncertainty.
b. Multiply the fractional uncertainty by the power (e.g., two if quantity is
squared).
c. Calculate (absolute) uncertainty of this quantity.


















9

B. Sample Experiment with Uncertainties Parameters
Notes to teachers:
The data used in this experiment is extracted from a few sample experiments conducted for
Physics High Level (HL) under the International Baccalaureate Diploma programme. Teachers
are advised to read and understand thoroughly the materials presented. This section is
meant for teachers only and can be used as guide for teaching and conducting lab
experiment for Uncertainties in Measurement.

Aim of Experiment
To determine gravitational acceleration using a simple pendulum.
Introduction
A pendulum is a weight suspended from a pivot so it can swing freely. When a pendulum is
displaced from its resting equilibrium position, it is subject to a restoring force due to gravity
that will accelerate it back toward the equilibrium position. When released, the restoring
force will cause it to oscillate about the equilibrium position, swinging back and forth. The
time for one complete cycle, a left swing and a right swing, is called the period.
In this experiment, the length of the string will be altered to see the effect on the time period.
With the collected and calcula3333ted values, the value of acceleration due to gravity will be
determined as gravity plays a vital role in the motion of the pendulum (and the only force
acting).
The bob has to travel more distance in the given time. As the length of the string is being
changed, it is the independent variable. This would affect the time period, hence being the
dependant variable. Factors like the mass of the bob (154.02g 0.01g), angle of release (45
1), the stopwatch (0.01s).
The main purpose of the experiment is to find one factor that affects the time period of a
simple pendulum. In this case, the factor is length of the string.
Theory
A simple pendulum performs simple harmonic motion, i.e. its periodic motion is defined by
an acceleration that is proportional to its displacement and directed towards the centre of
motion. It can be shown that the period T of the swinging pendulum is proportional to the
square root of the length l of the pendulum:

10

g
l
T t 2 =
with T the period in seconds, l the length in metres and g the gravitational acceleration in
m/s
2
. Our raw data should give us a square-root relationship between the period and the
length. Furthermore, to find an accurate value for g, we will also graph T
2
versus the length
of the pendulum. This way, we will be able to obtain a straight-line graph, with a gradient
equal to 2 (g)
1
.
Hypothesis
As the length of the string decreases, the time period also decreases. This is because, as the
length of the string decreases, the bob has to travel less distance in the same time (10
oscillations). And as there is less distance to cover in the same number of oscillations, the
velocity of the pendulum increases which hence decreases the time period. This means that
the length of the string is proportional to the period.
( ) l (T) string of Length period o
Apparatus and Materials
Weighing scale ( 0.01 g)
Retort stand
Meter stick ( 0.05 cm)
Stopwatch ( 0.05s)
String
Pendulum bob ( 0.01g)
Protractor ( 1)
G clamp

11

Design

Figure 1: Diagram of the set-up for the experiment
Procedure
1. After all the materials are acquired, set up the apparatus as below.
2. Set the length of the pendulum to be 100 cm.
3. Set the bob till 45 degrees mark (as precisely as possible) and then leave it.
4. As soon as it is left, start the stopwatch and record the time for ten oscillations.
5. Repeat the step 2 until step 4 two more times without changing the length in
order to get three trials.
6. Now, decrease the length of the string by 10 cm and repeat the steps 2, 3, 4 and 5.




12

Data Table
The data table below shows the raw data collected which includes the decreases in the
length of the string and the time trials for ten oscillations.
Time for 10 oscillations, t/s
Trial 1 Trial 2 Trial 3
Average
time, t / s
0.05 cm 0.6 s 0.06 s
10 7.5 7.7 7.3 7.50 0.75 8.0
20 9.7 9.9 9.5 9.70 0.97 6.2
30 11.4 11.6 11.2 11.40 1.14 5.3
40 13.3 13.1 13.5 13.30 1.33 4.5
50 14.5 14.9 14.1 14.50 1.45 4.1
60 16.0 16.6 15.4 16.00 1.60 3.8
70 17.1 17.0 17.2 17.10 1.71 3.5
80 18.0 18.3 17.7 18.00 1.80 3.3
90 19.0 19.1 19.0 19.03 1.90 3.2
100 20.2 20.0 20.4 20.20 2.02 3.0
Length of
string, l /cm
Fractional
(%)
uncertainty
for T
Period,
T/ s

Data Processing
These raw data need to be processed:
Find the average of the three trials, which will reduce our error, stated in the header.
Find the largest spread to determine the uncertainty for the recorded time for 10
oscillations. This can be done by finding the average of a three trials of each length and
find the biggest different between the largest reading to the average value.
( )
s 0 . 16
s
3
0 . 48
3
s ) 4 . 15 6 . 16 0 . 16 (
3
3 Trial 2 Trial 1 Trial
cm 60 Average
=
=
+ +
=
+ +
= = l

s 6 . 0
s ) 0 . 16 6 . 16 (
value Average reading Largest Spread
=
=
=


13

Since the time measurement for the 50 metre length had the largest spread, 0.4 seconds,
and was therefore used as the uncertainty in the time measurement.
Find the period T (time for one swing).
s 06 . 0 60 . 1
s
10
6 . 0
10
0 . 16
10
s ) 6 . 0 0 . 16 (
n oscillatio complete 1 for Time
=
=

=

% 8 . 3
100
60 . 1
06 . 0
100
period
y uncertaint
y uncertaint of %
=
=
=
x
x

Since T is squared, the fractional uncertainty will be multiplied by 2 base on the
propagation rule of uncertainty.

0.05 cm 0.06 s
10 0.50 0.75 8.0 0.56 16.0 0.09
20 0.25 0.97 6.2 0.94 12.4 0.12
30 0.17 1.14 5.3 1.30 10.5 0.14
40 0.13 1.33 4.5 1.77 9.0 0.16
50 0.10 1.45 4.1 2.10 8.3 0.17
60 0.08 1.60 3.8 2.56 7.5 0.19
70 0.07 1.71 3.5 2.92 7.0 0.21
80 0.06 1.80 3.3 3.24 6.7 0.22
90 0.06 1.90 3.2 3.62 6.3 0.23
100 0.05 2.02 3.0 4.08 5.9 0.24
Length of
string, l /cm
Period,
T/ s
Fractional
(%)
uncertainty
for T
T
2
/ s
2
Fractional
(%)
uncertainty
for T
2
Absolute
uncertainty
Fractional
(%)
uncertainty
for l / cm




14

Based on the theory, to find an accurate value for g we have to obtain a straight-line
graph. This can be done by plotting graph T
2
versus the length of the pendulum. This way,
we will be able to get a gradient equal to
g
2
4t
.
Note:
The theoretical uncertainty in the length measurement would be 0.05 cm (metre ruler).
However, in the experimental set-up, the two end points (the one tied to the clamp, and
the one tied to the metal bob) gave rise to a bigger uncertainty, as the exact end-points
could not be precisely determined. We estimated the uncertainty in length to be 0.5 cm.



Conclusion and Evaluation
Graphing T
2
against l clearly shows a linear relationship, in agreement with the theory.
The actual line of best fit does not go through (0,0) which suggests a systematic error in
our experiment. The interception at
2
T axis is at 0.17. Using MS excel slope function;
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
0 20 40 60 80 100 120
T
2

/

s
2

l / cm
T
2
against l

15

we can calculate the gradient of the of the line of best fit which is cm s
2
0388 . 0 . Given
the linear equation of the line of best fit is:
( )
2 2 2
174 . 0 0388 . 0 s cm s l T + =
Our value for g than can be calculated by dividing 4
2
with the gradient.
2
2
2
2
2
17 . 10
88 . 3
478 . 39
01 . 0
0388 . 0
478 . 39
0388 . 0
4
s m
m s
m
s
cm s
g
=
=
|
|
.
|

\
|
=
=
t


figures t significan 1 to 2 . 0
17 . 0
100
50 . 16
100
50 . 0 0 . 16
100
for y uncertaint fractional for y uncertaint fractional
for y uncertaint
2
2
s m
l T
g
=
=
=
+
=
+
=


2
s m 2 . 0 17 . 10 = g
Looking at our graph, we cannot identify any outliers. However, our data values suggest
a line of best fit that does not pass through (0,0). When we do fit a linear regression onto
our data values, that passes (0,0), we see that the line does not hit all the horizontal
error bars (the uncertainty in the length). This may suggest a systematic error in the
measurement of the length of our pendulum.
Furthermore, this experiment had to be carried out in about one hour, with very basic
equipment. This, perhaps, contributed to the slight difference in the value for g that we
found.



16

Finding gradient of a slope in MS Excel
Returns the slope of the linear regression line through data points in known_y's and
known_x's. The slope is the vertical distance divided by the horizontal distance between
any two points on the line, which is the rate of change along the regression line.
Syntax: SLOPE(known_y's,known_x's)
The equation for the slope of the regression line is:


1
2
3
4
5
6
7
8

A B
Known y Known x
2 6
3 5
9 11
1 7
8 5
7 4
5 4
Formula Description (Result)
=SLOPE(A2:A8,B2:B8) Slope of the linear regression line through the data points above (0.305556)











17

REFERENCES
1. Bell, S. (2001). The National Physical Laboratory. Measurement Good Practice
Guide No. 11 (Issue 2). A Beginners Guide to Uncertainty of Measurement.
2. Stanbrough, J.L. (2008). Science, Measurements, Errors, and Uncertainty.
Retrieved from
http://www.batesville.k12.in.us/physics/APPhyNet/Measurement/Measure
ment_Intro.html
3. Uncertainties in Measurements. (20100. Retrieved from:
http://chemwiki.ucdavis.edu/Analytical_Chemistry/Quantifying_Nature/Unc
ertainties_in_Measurements
4. Errors and Statistics - Accuracy and Precision. EU Socrates Minerva
Programme. Retrieved from:
http://www.lepla.org/en/modules/Activities/p04/p04-error1.htm