You are on page 1of 7

Estimation of uncertainties in

measuremenaddasts anfgrdd
their subsequent
calculasaddastionssdasdedsfdxx
Introduction
Throughout this document the following conventions will be used:

refers to some experimental quantity that can be measured, e.g.

voltage or distance.

refers to the mean of a number

measurements of quantity

refers to the precision in the measurement of the quantity

x .
and is

also known as the uncertainty or error associated with that measurement.

x x

is the average of the measured values, give or take the precision of the

measurement.

refers to the number of times

refers to the standard deviation of a set of numbers

( x ix )2
n

was measured.

1
2

The standard error =

In experimental science measurement is a vital part of our attempts to observe


the world around us. No quantity can be measured exactly. That means we can
never record a value correct to an infinite number of decimal places. What we
can do is estimate what the value of our measurement is, and indicate the range
of values which set its upper and lower limits.
Accuracy in measurement is about:

1. Estimation of the true value of the measured quantity


2. The precision with which that measurement was made
When we report that, for example, the resistance of a wire was measured to be
2.5 0.1 , we are identifying that the best estimate possible with the
equipment used is 2.5 and that the precision of the measurement is to within
0.1 . We are 95% confident that the true value of the resistance is not greater
than 2.6 nor smaller than 2.4 .
In general throughout this text the expression

x x

is used as shorthand,

meaning the average of the measured values, give or take the precision of the measurement.
The precision of the instrument is not the precision of the measurement in all
cases. When multiple readings of a value are made, the measurement is not now
the result of a single act of using the instrument to obtain an estimate of the
value of the subject of investigation. Rather, the best estimate of that value is
the mean of all the readings recorded.
When we use an analogue instrument such as a ruler we need to estimate by
eye where the value is on the scale. Such instruments are prone to parallax error
by users in reading the scale. Even a digital instrument has an element of error
associated with its measurement it is a mistake to think that the number
displayed on the screen is exact, it is not. The number displayed represents a
range of voltages between two specified measured values, and electrical devices
themselves have material characteristics that determine how they physically
respond to the presence of electric current.
When we measure a quantity we have to quote our best estimate of the
measured value and also we have to quote the precision of the measurement.
This second value is also known as the error or the uncertainty in the measurement
and really indicates that we are confident that the value we have measured lies
between certain identifiable limits.
The standard deviation is a measure of the spread of a statistically normal
population. Given that in a well-defined experiment the only sources of
unaccounted for error will be random in their nature, any set of measurements
will display a normal distribution. That being the case the standard deviation of a
set of measurements is a valid starting place to establish an estimate of the
precision of a certain class of measurements. A useful quantity is the standard
error:

x =

n ,

Where denotes the standard deviation of the data and n is the number of
measurements made.
There are three situations that need to be accounted for

1. The uncertainty associated with a single measurement


2. The uncertainty in the mean of a number of repeat measurements of a
single quantity
3. The uncertainty in a quantity that is a function of other quantities, each of
which have their own uncertainties.

The uncertainty associated with a single measurement


We measure

x . We need to be able to quote

x x .

One measurement is not able to be analysed statistically, so such concepts are


no help here.
When one measurement is made, the common practice is to quote the value to
the nearest number on the scale (or screen if digital) as the best estimate of the
value of the measured quantity and to use half the instrument resolution as the
estimate of the precision of the instrument.
For example, if the length of a pencil were to be measured using a ruler marked
in mm graduations, and we found that length to be 65 mm, we would quote that
measurement as 65 0.5 mm. We are acknowledging the fact that we dont
know that the pencil is exactly 65.0 mm long and not 0.5 mm longer or shorter
as we cannot measure more precisely with that instrument.

Significance of uncertainties
It becomes possible to make very accurate measurements only when the
precision of the instrument is not of comparable size to the object being
measured. To estimate the significance of the uncertainty in a measurement,
divide the uncertainty by measured value and then multiply by 100. To measure
the length of a grain of rice to be 2.5 0.5 mm is an uncertainty of 20%.

uncertaintythe measurement
100
best estimate of themeasurement

e.g.

0.5
100=20
2.5
so the uncertainty is 20% of the value of the measurement, which is getting to
be quite a lot.
Another example is the use of an alcohol thermometer to measure the
temperature of a small volume of liquid. If the volume of the liquid to be
measured is similar to the volume of the bulb of the thermometer, then the heat

exchange between the sample and the measuring device will be such that
making the measurement will alter the measurement by up to 100%, rendering
the measurement useless. Ideally, the act of measuring the temperature should
not alter the temperature being measured, so we should ensure that the mass of
the thermometer is very much smaller than the mass whose temperature we
wish to measure in order to reduce that effect.
An example familiar to students involves the use of a stopwatch to time a falling
object as it falls through a short distance (1 metre, for example). The duration of
the event is actually of the same order as the reaction time of the human
operator squeezing the stopwatch. Accurate estimation of such short durations
( approximately 0.1 seconds) is impossible by stopwatch.

Digital instruments
When using digital instruments it is a mistake to assume that the number on the
screen is exact. The rule here is that you use half of the smallest unit displayed.
If a volt meter shows a reading of 2.00 volts, then the measurement is (2.00
0.005) volts.
The only reason one might have to quote a differing value would be if such was
supplied by the manufacturer of the device at the time of purchase.

The uncertainty associated with multiple measurements


of the same quantity
When we repeat measurements the correct and obvious thing to do is to
calculate the mean value and use that as the best estimate of the measured
value. The uncertainty in that value cannot be obtained from the instrument
itself. Instead we can use the set of numbers that result from our multiple
measurements to determine the quality of those measurements.
Use of a spread sheet application such as Excel makes these calculations very
easy to do, as one can simply apply the standard deviation function to any set of
numbers one they have been entered into a column or row.
Determination of the standard error is then a matter of combining the standard
deviation and square root functions. The table below contains data obtained by a
student who had been given the task of determination of the volume of a table
top. Ten repeat measurements were made of each dimension. With ten
measurements it is possible to calculate the mean and standard deviation and
subsequently to use those values in determination of the uncertainty of the final
quantity. With each set of measurements the uncertainty is calculated as the
standard error,

n .

Trial
1
2
3
4
5
6
7
8
9
10
ave
std. dev.
std. error
% error
sum of % error
volume estimate
(cm3)

Width
(cm)
119.0
119.1
119.4
119.1
119.3
119.0
119.2
119.0
119.2
118.9
119.1
0.15
0.05
0.04%
0.73%
13222.1

Length
(cm)
59.9
59.8
59.8
60.0
60.0
59.7
60.0
59.7
60.0
59.7
59.8
0.12
0.04
0.06%

Depth
(cm)
1.8
1.9
1.9
1.9
1.9
1.8
1.9
1.9
1.9
1.9
1.9
0.04
0.01
0.63%

error in volume
(cm3)

96.5

Table 1: measurements of the dimensions of a tabletop, combined to determine the


volume of the tabletop

Combining Uncertainties
Combination of uncertainties can be an involved process, but there are
defensible techniques that a high school student can employ.
Take as an example a case where the object was to determine the area of a
rectangle, and two single measurements have been made only. Then the
uncertainty in each is half the smallest graduation of the instrument scale. Each
of these is converted into a percentage. The percentages can be added and then
the resultant sum is used to determine the total uncertainty. If the length was
measured to be 100 0.5 mm and the width to be 50 0.5 mm, then the
uncertainties in each will differ.

0.5
100=0.5 , and
100

0.5
100=1 , so the total uncertainty in percentage
50

terms is 1.5%.
To estimate the area use the product of the two measured values:

A=l w=100 50=5000 mm

1.5% of 5000 is 75 so the value for the area we quote is (5000 75) mm2
The table top exercise furnishes us with an example involving multiple
measurements. The standard error in the mean value of each dimension of the
table top is converted into a percentage. The percentages can be added together
directly. The final percentage is then applied to the estimated volume of the
table top, obtained from the product of the mean values of each dimension (i.e.

h w d ). From the example data we calculate that the best estimate of the
volume of the table is 13, 222.1 cm 3, and that there is 0.73% uncertainty in this
value, which in the same units as the mean is 96.5 cm 3.
The value that one should quote as the volume of the table is (13 222.1 96.5)
cm3.

You might also like