Sei sulla pagina 1di 17

How reliable is your data?

- Uncertainty in Chemistry
Introduction
All experimental data is limited by the method of measuring that quantity and all
measurements that you make in the lab have some degree of uncertainty
associated with them.
(Uncertainty is often referred as error - well be using the terms
interchangeably.)
The number of significant figures used to record experimental data is very
important.
In Maths, 10 is the same as10.00.
However in an experimental science where we measure and gather data, the
situation is quite different:
When an experimental value is reported, the number
of significant figures gives the reader an indication
of the reliability (or precision) of that quantity.
Significant figures
When we take a reading from, say, a ruler or a burette, the value we note
consists of digits we are sure about plus one more which we estimate.
The following diagram, taken from the excellent website1 of Martin
Hackman of the Physics Department, Idaho State University, illustrates
this point:

Using the ruler on the left, we would note down the length of the grey bar
as 3.85 cm.
We are sure about the 3 and the 8 but the 5 is just an estimate
Using the ruler on the right, the length of the grey bar would be recorded
as 3.8 cm. We are sure about the 3 but are estimating the 8.
When a measurement is quoted, the number of significant figures is the
number of digits you are sure about plus one more.
So the 3.85 cm measurement has three significant figures and the 3.8 cm
measurement has only two sig fig.
Similarly, if you record a burette reading as 2.5 mL, this quantity has 2 sig
fig and you are claiming that youre certain about the 2 mL but not
entirely sure about the 0.5 mL part.

http://www.physics.isu.edu/~hackmart/sophlbin.pdf

NB: The number of decimal places is NOT the same thing as the
number of significant figures.
For example:
A measurement of 1035 has four sig fig youre confident about the 1 and
0 and the 3 digits but not the 5.
1.23 has three sig fig.
23.48 x 10-3 has four sig fig.
1.2340 has five sig fig.

Significant figures (continued)


When looking at significant figures,
we do NOT include left-hand side place-holder zeroes:
0.00348 has three sig fig.
0.01054 has four sig fig.
If there are zeroes on the right hand side of the number then it is
difficult to decide whether these are simply place-holders (and
therefore not to be counted) or whether they are indeed bona-fide
contributors to the number.
However, in some cases, you just dont know the number of significant
figures.
Here are some examples:
1.00 definitely has three sig fig since neither of the zeroes needs
to be there as a place holder.
100 has an indeterminate number of sig fig since we do not know
whether the two zeroes are bona-fide measurements or simply
place-holders which put the 1 into the hundreds column.
On the other hand,
1.00 x 102 certainly has three sig fig and
1 x 102 certainly has only one sig fig

Test yourself:
Give the number of significant figures in each of the following experimental
values.
If the number of sig fig is indeterminate, say so.
a) 0.00008

b) 1.0235 x 1023

c) 0.0104

d) 1065

e) 1000

f) 1053

g) 1.000 x 103

Reporting uncertainty
The uncertainty associated with a measurement can be depend on:
your personal skills or technique (or lack thereof);
the equipment or instruments involved;
3

the actual method of measurement.


You need to be able to inform the reader of your report about the
uncertainty associated with your measurements. This is done in two ways.
When you report an experimental value:
the number of significant figures should reflect the reliability
(or precision) of that quantity.
the experimental value should be accompanied by a +/figure showing the uncertainty associated with that value.

Reporting uncertainty (continued)


For example, lets say I report a mass of 8.6753 0.0002g.
This mass has been reported to five significant figures. It would have
been measured on an analytical balance capable of weighing to four
decimal places.
The 0.0002 part of this data is the uncertainty or error
associated with the reported mass.
The 8.6753 means that I am confident about the 8,6,7 and 5 but not
about the final 3.
The reported uncertainty of 0.0002 supports that interpretation and
tells the reader that the mass of the item should be somewhere
between
8.6751 g and 8.6755 g and that if I measured it again, I would obtain a
value in that range.

When you report a final value,


the uncertainty value is usually rounded off to one sig fig

and then

the reported quantity is rounded off to the same number of decimal


places as the uncertainty.

For example:
Lets say a final mass was calculated to be 2.089 g and the
uncertainty associated with that final mass was calculated to be
0.0213 g.
Reporting this mass as 2.089 0.0213 g is not useful because the 2nd
decimal place (2.089 0.0213) is the spot where the largest uncertainty
occurs.
In other words - if we're uncertain about the value of the 2nd decimal
place of our mass, there is not much point in reporting the much smaller
uncertainty in the 4th..
Therefore we round off to the 2nd decimal place and report the mass
as 2.09 0.02 g

Test yourself:
Report the following results and uncertainties appropriately:

a) 1.08512 0.01923 mL
b) 100.345 11.1g
c) 0.009832 0.00096 g
d) 13.00234 0.00145 g
e) 1067 107 kg
f) 123.17 0.39 mL

Absolute and Relative Uncertainty


The uncertainty associated with a reported value can be expressed as
either absolute or relative uncertainty.
Absolute uncertainty
has the same units as the measured quantity (2.09 0.02 g)
Relative uncertainties
are often used when you are calculating uncertainties (see Error
Propagation below) or when you are comparing or manipulating
uncertainties of quite different measured properties.
The relative uncertainty is calculated by expressing the absolute
uncertainty as a fraction of the measured value.
This fraction can be expressed as a decimal fraction without units or as a
%.
Example:
The 0.02g absolute uncertainty in the 2.09g mass mentioned above
corresponds to a relative uncertainty of 0.02g/2.09g = 0.001 or 0.1%
Example:
I have measured the mass of a strange sort of sample and found it to
be 101g.
I have also reported that its volume is 92 1mL.
Intuitively, the volume measurement of 92 1mL is more precise (has
less uncertainty) than the mass measurement of 10 1g.
This can be demonstrated by comparing the relative uncertainties of
the two measurements. The relative uncertainty associated with the
volume measurement (%uncertainty = 1mL ! 100 = 1.1% ) is much less
92 mL

than that associated with the mass (% uncertainty =

1g
! 100 = 10% ).
10g

Test Yourself:
Convert the following absolute uncertainties to relative uncertainties.
Round-off the relative uncertainty to one significant figure.
Q1 a) 201mL
b) 2822mg
c) 0.01230.006g
d) 2.250.2L

Q2 Express the following relative uncertainties as absolute


uncertainties
Round off and report the final reported quantity and its absolute
uncertainty appropriately.
a) measured quantity is 20.037mL. Rel uncert = 0.05
b) measured quantity is 6.4571g. Rel uncert = 0.1%
c) measured quantity is 250.7mL. Rel uncert = 0.005
d) measured quantity is 0.3469kg. Rel uncert = 5%

Calculating Uncertainty: Error Propagation


Sometimes the quantity you report is the result of a single observation.
For example, you might use a bulb pipette to deliver a set volume of liquid.
(This set volume is called an aliquot).
In a case like this, reporting the error associated with that volume is
simple because, provided you use the pipette properly, you can quote
the value etched on the surface of the pipette by the manufacturer
since the manufacturer guarantees the delivered volume to be within
the etched value.
Alternatively, you could calibrate the pipette yourself and work out your
own uncertainty. This second approach would include uncertainty
associated with your pipetting technique.
However, the quantities that you provide at the end of a report are usually
obtained from calculations which use a number of different measurements.
These measurements may have been added, subtracted, multiplied or
divided together during your calculations.
You need to be able to determine the uncertainty of each measurement and
use these uncertainties to calculate the uncertainty of your final calculated
answer.
The uncertainty or error associated with your final answer ALWAYS
increases with the number of measurements used during the calculations
leading to that final answer.
This inexorable growth of uncertainty is called error propagation.
The basic rules for calculating uncertainties are:
When you add or subtract quantities,
ADD THEIR ABSOLUTE UNCERTAINTIES
When you multiply or divide quantities,
ADD THEIR RELATIVE UNCERTAINTIES.
Some examples are shown on following pages.

Calculating Uncertainty: Error Propagation (continued)

Example 1
Simple example using the mass and volume values of that strange sample
from p. 20
density

mass
volume

0.1087 g.mL-1

10 g
92 mL

Since volume has been divided into the mass, add their relative uncertainties:
Total % uncertainty
in density

= rel uncertainty in mass + rel uncert in volume


1g
1mL
=
+
! 100
! 100
10g

10%

11.1%

92 mL

1.1%

Therefore the absolute uncertainty of the density (i.e. the bit) is 11.1% of
the density
abs uncertainty = 11.1% of 0.1087 g.mL-1
= 11.1/100 x 0.1087 g.mL-1
= 0.012 g.mL-1
So, before we round off:

NB: We could have


expressed the relative
uncertainty as a
decimal fraction
rather than as a %.
The final answer
comes out the same.

density = 0.1087 0.012 g.mL-1


These numbers show that we are uncertain about the 2nd decimal
place; so we are only entitled to quote our result to that many decimal
places.
As noted earlier, since the uncertainty is just an estimate, it is usually
rounded off to one significant figure.
So we would round off 0.012 g.mL-1 to 0.01 g.mL-1
(there is no point in reporting the third decimal place if there is
uncertainty in the 2nd).
The density value itself is then also rounded off to the same decimal
place as the uncertainty ie 0.1087 g.mL-1is rounded off to 0.11 g.mL-1
After rounding off, we quote the density as 0.11 0.01 g.mL-1

10

Example 2 Using an analytical balance:


On our analytical balances, the error in each reading is 0.0001
g.
When you weigh something on a balance, you actually take two
measurements because you always tare the balance first ie set it to
zero. This is the first reading: 0.0000 0.0001 g even though you
never write it down.
When you have placed your sample of the balance, the second reading is
the number you actually record and it might be, say, 6.5432 0.0001 g.
The mass of the item is the difference between the two readings.
Because the data are subtracted, you can add the absolute
uncertainties:
second reading
first reading
mass of item

6.5432 0.0001 g
0.0000 0.0001 g
6.5432 0.0002 g

So although every reading has an uncertainty of 0.0001 g,


every mass you obtain directly from the balance has an uncertainty of
0.0002 g.

Example 3 - Taking a reading from a burette.


Your demonstrator says
you should be able to take a reading to the nearest 0.01 mL from the burette
(see using a burette in the Lab section of this manual).
You report that the bottom of the meniscus is at 12.72 mL.
So youre sure about the 12.7 but are just making an intelligent estimate
of the final 2.
If you write the uncertainty of this reading as 0.01 mL,
then you are claiming that you are sure that the true reading is
somewhere between 12.71 and 12.73 mL.
This is OK for most people, but if your eyesight is bad or if for some
reason the meniscus line is unclear, you may prefer to play it safe and
claim an uncertainty of 0.02 mL (or more) for each reading of the burette.

Example 4 - Reporting a volume delivered using a burette.


The volume delivered by a burette is calculated from two readings.
11

You subtract the first reading (12.72 mL) from the second (22.82 mL)
to get the volume delivered (10.10 mL).
When quantities are added or subtracted,
ADD THE ABSOLUTE UNCERTAINTIES.
If your eyesight is good (see Example 2), each reading has an error of 0.01 mL
So the error in the delivered volume is 0.02 mL.
Therefore you would report the delivered volume as 10.100.02 mL.

12

Example 5 - Reporting the mass of sodium carbonate added to a 500 mL flask.


In this example, the added mass was worked out using the difference method:
First you recorded the mass of the weighing bottle plus sodium carbonate and
noted down 13.1234 0.0002 g .
Then you tipped the sodium carbonate into the flask (no spills)
Then you weighed the (almost) empty weighing bottle: 10.1134 0.0002 g.
The difference is the mass of sodium carbonate added to the flask is reported as
3.0100 0.0004 g.
(Remember that you add the absolute errors
whenever you subtract or add measured quantities).

With more lengthy sets of calculations which multiply or divide a series of


measured quantities, you can add the relative uncertainties all in one go a
table is often useful for keeping track of quantities and their absolute and
relative uncertainties.
An example of such a table is shown below.
Example 6 - calculating the concentration of a solution of sodium carbonate.
You added 3.0100 0.0004 g of sodium carbonate to a 500 mL
volumetric flask and made up the volume in the flask to the mark on the
neck of the flask.
Since the etching on the neck says that the uncertainty in flask volume is
0.2 mL,
the volume in the flask should be reported as 500.0 0.2 mL.
The molar mass of the sodium carbonate was given as 105.989 g.mol-1
conc =

moles
mass
3.0100 g
-1
=
=
= 0 .056798 mol.L
vol in litres molar mass x vol in litres 105.989 g.mol-1x 500.0 x 10 -3 L

quantity

abs
uncertainty

relative
uncertainty

0.0004 g

0.00013

mass

3.0100 g

volume

500.0 mL

molar mass

105.989 g.mol-1

insignificant

insignificant

concentration

0.056798 mol.L-1

0.00003 mol.L-1

0.00053

0.2 mL

0.0004

13

concentration
0.056800.00003 mol.L-1
rounded off

14

Accuracy and Precision


When you are making a measurement, you often do it two (duplicate) or
three (triplicate) or even more times.
Accuracy is how close your average measurement is to the correct or
real value.
Often, in real life, we dont know the real value (thats why were
measuring it in the first place) but we often have some rough idea of
what to expect and use that knowledge to set up the measurement
protocol in the first place.
In our laboratory exercises, the staff have a pretty good idea of what the
real value should be since a very experienced and capable laboratory
technician has set up and done the experiment beforehand. We can
therefore tell how accurate your results are.
Precision refers to the extent to which your measurement can be
repeated with the same result. In other words, precision refers to the
reproducibility of your measurement.
The greater the spread of your measurements, the poorer your
precision.
Poor precision usually indicates that you need to improve your
experimental technique or the equipment is faulty or both.
Think of the true value of your measured quantity as the centre of
a bulls-eye.
If all your measurement arrows hit smack in the middle of the
bulls-eye, you have good precision and accuracy.
If all your arrows hit very close together but not in the centre,
then you have good precision but poor accuracy.
If all your arrows are scattered over the target, then you have
poor precision.
In this last case, the accuracy is also regarded as being poor
even if your average is really close to the true value - since a
spot-on average value, accompanied by poor precision, is
simply due to luck.
When a value is measured a number of times, you should report the
average (mean) value.
The uncertainty associated with the mean value will be the
inherent uncertainty of each measured value or half the range of
the measured values, whichever is bigger.
Example:
15

Lets say youve weighed four aliquots1 delivered by a pipette.


Each mass should be very similar (if your technique is good and the pipette is not
faulty).
An analytical balance was used and the uncertainty of each mass is 0.0004 g.
The masses of the aliquots are:
10.2345 0.0004 g
10.2360 0.0004 g
10.2352 0.0004 g
10.2340 0.0004 g
The mean mass is
10.2349 g
and the range =
largest measured value smallest measured value
=
10.2360 g -10.2340 g
=
0.0020 g
So 1/2 range is 0.0010g.

an aliquot is the fixed volume of liquid delivered by a bulb pipette

16

Since this half range (0.0010g) is greater than the intrinsic uncertainty of each
mass measurement (0.0004g), we report the uncertainty as 0.001g.
(Note that the final uncertainty value is only quoted to one sig fig.)
The mean mass of your four aliquots would therefore be recorded as 10.235
0.001g
(note the rounding off to the third decimal place).

Your neighbour (who uses a pipette better than you or is just really
lucky) did the same experiment and obtained four identical masses of
10.2350g!
Although 1/2 range in this case is 0.0000g, the uncertainty would be
reported as 0.0004g which is the inherent uncertainty associated with
the mass of an aliquot.

17

Potrebbero piacerti anche