You are on page 1of 5

Accuracy, Precision and Standard Deviation



The following sections deal with the ideas of precision and accuracy in greater detail and describe the use of standard deviation in evaluating analytical data.


Standard deviation and precision

The greater the degree of scatter of a set of measurements the lower is the precision of those measurements. Precision is often measured by the standard deviation of the set. The standard deviation s of a set of n repeat measurements is defined as
s (x n x )2 ¯ 1

where x is a single measurement and x is the mean (average) measurement. The ¯ symbol means ‘sum of’. The lower a standard deviation of a set of repeat measurements, the better is the precision of those measurements. The relative precision of two or more methods of measurement (e.g. of two analytical methods) is compared by calculating their percent relative standard deviation. The percent relative standard deviation (%RSD) is calculated from the standard deviation s and mean measurement x using the equation: ¯
%RSD = 100 x ¯ s

The standard deviation enables us to estimate the spread of measurements on either side of the mean measurement. If the only errors in the measurements are random ones, there is a 90% chance that the mean value of a measured quantity will be within the confidence limits
x ¯ 2 s n

However, to rely on such a calculation we need to be sure of our value of s: if the spread of measurements is considerable, we may have to make 10 or more repeat measurements.



04 0.27 0.0 5.5 5. there is a 90% chance that the mean benzene concentration is in the range 5.23/5. the following Microsoft Excel spreadsheet shows calculations of the mean.3.3 (2 0. standard deviation and %RSD of the five numbers in column B: .2% We conclude that provided that systematic errors are absent and that we are confident we have taken sufficient measurements.00 0.3 5. 5.3 0. Using a calculator or a spreadsheet to calculate standard deviations 3 It is very useful to be able to calculate standard deviations using your pocket calculator: study the calculator instructions – or ask a knowledgeable friend! A spreadsheet.3 6 5.2 APPENDIX 1 Example 1 Repeat measurements of the percentage by volume of benzene in a fuel gave the following results: 5.09 0. (ii) Range of the mean concentration s range x ¯ 2 5.5.5–5.5 x ) (x ¯ 5.0.3% 0.01 0.09 0.1%. To illustrate this.2 and 5. (iii) Calculate the %RSD of the measurement Answer (i) Standard deviation mean x 5. 5.3 4.0 5.45)) n 5. 5. (iii) %RSD 100 x ¯ s 100 0.0.5. 5.0 5.2 0.04 0.5 x ¯ (x (5.3%. (ii) Estimate the confidence limits of the mean concentration.2 5. such as Microsoft Excel. may also be used to carry out the calculations.27 —— 5 0.0 0.23/(2.27 ——— n 1 0.0 x )2 ¯ 5.5) 5. (i) Calculate the standard deviation of the results.23% s (x x 2 ¯) ———— — n 1 Note that the units of the standard deviation are the same as that of the quantity being measured – here the percentage.2 5.3 0.1 0.3 0.2 sum 0.

74 3.211 1.2215.2210 and 1.77 .2214.43 2.303 1. Example 2 The concentration of magnesium in a tank of water was found using two different analytical methods. 1. only rows 1–8 and columns A and B are shown. 1. The results were: Sample 1 2 3 4 5 AAS EDTA Titration [Mg2+]/mg dm–3 2. 1.2219.31 2. and five of these were used for five repeat measurements (after appropriate dilution) using AAS and the remaining five used for repeat measurements using the EDTA titration. 3) =ROUND((B7/B6)*100.241578 16.82 3. Calculate the percent relative standard deviation and the confidence limits of the mean of the concentration.50 2.75 3.2217. Exercise 1A The concentration of chromium in a powder was measured six times.435 1.52 2. (I) atomic absorption spectroscopy (ASS) and (II) by titration with EDTA.79 3.ACCURACY.2208. Twenty 10 cm3 samples were withdrawn from the tank.4496 0.66513 For simplicity. PRECISION AND STANDARD DEVIATION A 1 2 3 4 5 6 7 8 mean Std deviation %RSD 3 B 1.843 1.456 1. B7 and B8 were =AVERAGE(B1:B5) =STDEV(B1:B5) =(B7/B6)*100 If the following commands are used: =ROUND(AVERAGE(B1:B5). The commands entered in cells B6. The results (in units of g dm–3) were: 1.61 3. 3) =ROUND(STDEV(B1:B5). 1. 3) the answers are displayed to fewer significant places.

the AAS results are more accurate. symbolised ppm(v)).11/2.) General revision questions for Unit 1 1.9 2. . What are the units of k? 2. was measured at a traffic junction.4 3. s = 0. s = 0. Revision questions for Appendix 1 1.24.22. the percent relative deviation of both methods and the error of both methods.50 mg dm-3. 0.5 2.032 and %RDS = 0. (i) Calculate the mean concentration determined by each method and the standard deviation and %RSD of the results for each method.78 mg dm–3.47 mg dm–3) is closer to the true concentration and the error is only 2. 0.21. Repeat measurements over a five-minute period gave the following concentrations: 0.4 3.85 %.2 The true concentration of benzene in the flask is 3. 0. expressing your answer to the correct number of significant figures. and (ii) the confidence limits of the mean concentration of ozone. Calculate (i) the mean ozone concentration to two significant figures and the standard deviation to one significant figure.100 ppb. [NaOH] is the concentration of sodium hydroxide (also in mol dm–3) and k is a constant called the rate constant. Calculate the mean concentration of benzene determined by two methods. 0. Answer (i) For AAS.47 = 4. 2. The rate (‘speed’) of the reaction between bromopropane and sodium hydroxide is given by the equation rate of reaction = k [C3H7Br] [NaOH] where [C3H7Br] is the concentration of bromopropane (in mol dm–3).03 mg dm–3. 0. [Mg2+]mean = 3.78 = 0. in parts per million by volume.22. The results of four repeat measurements taken at the same time were: Measurement 1 2 3 4 Method A Method B Hydrocarbon concentration/ppb 2. (For example.47 = 0. Calculate the pH of the solution.32/3. Comment on the accuracy and precision of both methods. The poor accuracy of the results obtained by the EDTA titration suggest that the titration is subject to chemical interference. The ozone concentration in cm3 of ozone per million cm3 of air (i. The results from the titration have the lower RSD so that they are more precise (less scattered) than those obtained by AAS.50 – 2.0545 mol dm–3.47 mg dm–3. The concentration of hydrogen ion in a solution is 0.e. 0. On the other hand.21. For titration.5 %. 0.3 2. [Mg2+]mean = 2.5 2. 0.11 and %RDS = 0. (ii) The best way to assess the precision of the methods is to compare their %RSDs.23 and 0.19.24. (ii) Comment upon the accuracy and precision of each method.22.23. this might be because other metal ions present in the mixture are combining with the EDTA.6 3. The units of rate of reaction are mol dm–3 s–1. A and B. since the mean AAS concentration (2.4 APPENDIX 1 Example 2 continued The true concentration of magnesium in the tank water is 2. The concentration of benzene in a sealed flask was determined by two analytical methods.

0957 %RSD 3.225 – 3.100) = 0.22 0.625 ppb (for Method A) and : (3. A Mean 2. PRECISION AND STANDARD DEVIATION 5 Answers Exercise 1A mean conc = 1. in the absence of systematic errors there is a 90% chance that the concentration of Cr lies between the limits 1.75 ¯ x 2 Revision questions for Unit 1 1. s 0.475 STD Dev 0.00034.22 ppm(v). the %RSD of method A is only 3.0341% confidence limits = 1.89%.2636 = 0.264.2217 – 1. However. 2001.22138/0.545) = –(–) 0.2210 g dm–3 Revision questions for Appendix 1 1. The errors are: (3.250 7.22138 ± 2 0.00417 g dm–3 %RSD = 100 (1.02 ppm(v).22138 g dm–3 Std dev = 0. Mean concentration s/ n 0. pH = –log[H+(aq)] = –log (0. 6 Therefore. so that Method B is more accurate than method A.225 0.013 2.00417) = 0.22138 ± 0.100 – 2.ACCURACY.21 ppm (v). k = rate of reaction/[C3H7Br] [NaOH] so that the units of k = mol dm–3 s–1/(mol dm–3)2 = mol–1 dm+3 s–1. Confidence limits 0.475) = 0. © Rob Lewis and Wynne Evans. so that it is the more precise of the two methods. Palgrave. B 3. Chemistry.000417 = 1.2636 = 0.125 ppb.23 0. 2.89 0. .