Food Analysis – FScN 4312W

Laboratory: Assessment of Accuracy and Precision

Key to Questions
1. Theoretically, how are standard deviation, coefficient of variation, mean, percent relative
error, and 95% confidence interval affected by: (1) more replicates, and (2) a larger size of
measurement? Was this evident in looking at the actual results obtained using the volumetric
pipettes and the buret, with n = 3 vs n = 6, and with 1mL vs 10mL?
Ans: (1) As the sample size increases (more replicates)

- Calculated mean approaches the true mean

- Standard deviation is inversely proportional to the square root of sample size, hence it
decreases

- CV decreases, as standard deviation approaches 0 for larger sample size

- % Relative error decreases as calculated mean approaches true mean

- 95% confidence interval narrows down

(2) Larger size of measurement

- Mean is close to true mean

- Standard deviation might be larger due to larger size

- CV decreases

- % Relative error decreases, as calculated mean is close to true mean

- 95% confidence interval narrows down

Not all of the above assumptions were confirmed in this experiment according to the data
observed. One reason could be that the sample size was very small (n = 3 and n = 6), to notice
significant differences, another reason could be human error and variation from one person’s
measuring technique to another.
2. Why are percent relative error and coefficient of variation used to compare the accuracy and
precision, respectively, of the volumes from pipetting/dispensing 1 and 10mL with the
volumetric pipettes and buret in parts 2 and 3, rather than simply the mean and standard
deviation, respectively?
Ans: Mean calculated from the readings gives the calculated mean, which may differ from the
true mean. Although good precision is obtained, accuracy is not guaranteed. On the other hand,
percent relative error calculates the error in the calculated mean with respect to the true mean.
Hence, percent relative error gives more information about how close the calculated mean is to
the true mean.
Co-efficient of variation (C.V) provides information about how much the standard deviation is
different form the calculated mean in terms of percentages. For example, if mean = 2 and
standard deviation = 0.1, then C.V = 0.05*100 = 5%. Therefore, we can infer that standard
deviation is 5% off from the calculate mean. Depending on the coefficient of variation we can
reject the data. Usually, 5% is the cutoff for accepting the data. But, we cannot vaguely reject the
data based on standard deviation. Standard deviation can be large for large measurements and
relatively small for small measurements. We cannot reject the data on the fact that standard
deviation is large for large measurements as compared to small measurements, unless C.V is
calculated. Hence, C.V is more useful parameter in data analysis as compared to standard
deviation.
3. Compare and discuss the accuracy and the precision of the volumes for the 1mL
pipetted/dispensed using a volume pipette, buret, and mechanical pipettor (parts 2, 3, and 4). Are
these results consistent with that what would be expected?
Volumetric pipette had more accuracy as compared to buret and mechanical pipette. Also, as the
sample size increased accuracy also increased for volumetric pipette and buret, but there was no
large change in the accuracy of mechanical pipette. This does not comply with what is expected
as mechanical pipette is supposed to be more accurate than volumetric pipette and buret. Perhaps
the mechanical pipette needed calibration.
On the other hand, the precision of mechanical pipette is better than buret and volumetric pipette
which is expected. This can be attributed to the fact that, human error in mechanical pipette is
much less. Generally, precision increases as sample size increases.
4. If accuracy and / or precision using the mechanical pipettor are less than should be expected,
what would you do to improve its accuracy and precision?
Ans: - Follow the instructions of how to use a mechanical pipettor closely

- Calibrate
5. In a titration experiment using a buret, would you expect to use much less than a 10mL
volume in each titration? Would you expect your accuracy and precision to be better using a
10mL buret or a 50mL buret? Why?
Ans: Large volumes are associated with small errors, i.e. the error due to 1mL in 10mL titration
is 10%, whereas the error due to the same 1mL in 100mL titration is 1%. Thus, we can reduce
the magnitude of error by increasing the measuring capacity. Thus, it is best to use diluted
reagents and therefore need a larger volume of the titrant to reach the end point.
However, when less than 10 mL is needed to reach the end point it is recommended to use 10mL
buret as it is more graduated than a 50mL buret. Thus, small amounts can be dispensed more
accurately from a 10mL buret as compared to a 50mL buret.
6. How do your results from part #5 of this lab differentiate “to contain” and “to deliver”? Is a
volumetric flask “to content” or “to deliver”? Which is a volumetric pipette?
Ans: Glassware is termed “to deliver” if it can dispense the entire amount of solvent it contains,
ex: buret. Whereas, some glassware can only store solvents, upon transfer of the solvent from
this glassware to another, the entire amount is not transferred.

Volumetric flask is “to contain” and volumetric pipette is “to deliver”.

7. From your results from part #6 of this lab, would you now assume that since a balance reads a
0.01 g that it is accurate to 0.01 g?

Ans: Not really. One balance can read 50 g standard weight as 50.02 g and another read 49.93 g.
The error may be due to instrument and hence cannot be avoided. Not to forget also that the last
figure might have been rounded up. Also, the balances have to be regularly calibrated.

8. What sources of error (human and instrumental) were evident or possible in parts #2-4, and
how could these be eliminated or reduced? Explain?

Ans: Systematic errors: This error explains the difference between precision and accuracy. The
results will be consistent but away from the true value. In other words, the results would be
precise but won’t be accurate. Identifying this type of error is very difficult as they are related to
faulty instruments, or non-calibrated equipment. Sometimes, human error can also lead to
systematic error; one example is that of buret. The proper way of taking readings from a buret is
to make sure that the bottom of the meniscus of the liquid is touching the top of the line you wish
to measure. If this rule is not followed every titration will be associated with an error of 0.2mL,
although consistent results are obtained. Other example is parallax error, which can be avoided
by ensuring that viewer's eyes are at the level of the graduation. These types of errors can also
be reduced by proper calibration of the instruments.

Random errors: Natural errors associated with instrument type or model, and all with each
experimenter’s techniques. These are impossible to avoid, but usually they are very small.

Blunders: Can be easily identified, as they are very obvious. These types of errors include using
a wrong reagent or instrument. This type of error is mainly caused by human negligence,
therefore proper practice and being careful will eliminate these kinds of errors.

9. You are considering adopting a new analytical method on your lab to measure the moisture
content of cereal products. How would you determine the precision of the new method and
compare it to the old method? How would you determine (or estimate) the accuracy of the new
method?

Ans: - Have multiple replications and then calculate and compare the standard
deviation/coefficient of variation and % relative error of the data produced by the old and
new method.
- Compare the data obtained from the new method with that of the data present in
literature (or) from data obtained by employing standard or official methods.

- Run a standard reference analysis using the new method and compare the result to that
of the previous method.