GIST 

Anal Chem 9/17/2016

3. EXPERIMENTAL ERROR
GIST Analytical Chemistry by Kim, Tae-Young

Why do we study this chapter?

2

Analytical Chemistry

Significant
Measurement Units
figures

Error
Systematic error
Random error
Propagation of errors

1
GIST Anal Chem 9/17/2016

1. Significant Figures
3

 Significant figures (SFs): The minimum number of digits required to express a value in
scientific notation without loss of precision.
 SFs = All of the certain digits + the first uncertain digits.
9.25 × 104 3 significant figures
9.250 × 104 4 significant figures
9.2500 × 104 5 significant figures
 Zeros are significant:
 They occur in the middle of a number or at the end of a number on the right-hand side of a decimal point.
106 0.0106 0.106 0.1060
 The last significant digit (farthest to the right) in a measured quantity always has some
associated uncertainty.
 By contrast, integers are EXACT!

2. Significant Figures in Arithmetic

: Addition & subtraction
4

 Rule of thumb: Rounding should only be done on the final answer

(NOT intermediate results), to avoid accumulating round-off errors.

 Addition & Subtraction: The result should have the same number of
decimal places as the number of with smallest number of decimal places.

 In case of numbers expressed in scientific notation:

 Express all numbers with the same exponent.
 Align all numbers with respect to the decimal point.
 Round off the answer according to the number of decimal
places in the number with the fewest decimal places.

2
GIST Anal Chem 9/17/2016

Multiplication & division

5

 Rule of thumb: Round off to the number of digits contained in the number with fewest
significant figures.

 The power of 10 has NO influence on the number of figures that should be retained.

 Rule for rounding off numbers

 If the digits beyond the last significant place is more than halfway to the next higher digit → Round up.
 If the digits beyond the last significant place is less than halfway to the next higher digit → Round down.
 If the number is exactly halfway (a 5) → Round to the nearest even digit. (Why?)
43.55 → 43. 6 43.45 → 43.4

Logarithms & antilogarithms

6

 Logarithm = Integer (Characteristic) + Decimal (Mantissa)

 Number of digits in mantissa of log x = number of significant figures in x:

 Number of digits in antilog x (= 10x) = number of significant figures in mantissa of x:

3
GIST Anal Chem 9/17/2016

3. Types of Error
: Systematic error
7

 Experimental error:“Uncertainty” associated with every measurement.

 There is no way to measure the “true” value of anything!

 Systematic error: Determinate error.

 Arises from a flaw in equipment or design of an experiment.
 If you conduct the experiment again exactly the same manner, the error is reproducible.
 In principle, this type of error can be discovered & corrected, although this may not be easy.
 Examples:
 An incorrectly standardized pH meter.
 An uncalibrated buret.

Experimental error
: Ways to detect systematic error
8

 Analyze a known sample, such as a certified reference material.

 Your method should reproduce the known answer.

 Analyze blank samples containing no analyte being sought.

 If you observe a nonzero result, your method responds to more than you intend.

 Use different analytical methods to measure the same quantity.

 If results do not agree, there is in one (or more) of the methods.

 Round robin experiment: Different people in several laboratories analyze identical samples
by the same or different methods.
 Disagreement beyond the estimated random error is systematic error.

4
GIST Anal Chem 9/17/2016

Experimental error
: Random error
9

 Random error: Indeterminate error.

 Arises from uncontrolled (and maybe uncontrollable) variables in the measurement.
 Has an equal chance of being positive or negative.
 Cannot be completely eliminated.
 Examples
 Subjective interpolation between the markings.
 Electrical noise in an instrument.

Precision vs. accuracy

10

 Precision: Reproducibility of a result.

 Accuracy: Nearness to the “true” value.

Bad precision Bad precision Good precision Good precision

Bad accuracy Good accuracy Good accuracy Bad accuracy

5
GIST Anal Chem 9/17/2016

Absolute & relative uncertainty

11

 Absolute uncertainty: The margin of uncertainty associated with a measurement.

 For a calibrated buret with the estimated uncertainty of ± 0.02 mL, ± 0.02 mL is the absolute uncertainty
associated with the reading.
 Relative uncertainty (dimensionless quotient):

 The relative uncertainty of a buret reading of 12.35 ± 0.02 mL,

0.02
0.002
12.35
 Percent relative uncertainty:

Propagation of uncertainty from random error

12

 For most experiments, we need to perform arithmetic operations on several numbers, each
of which has a random error.

 The most likely uncertainty in the result is NOT the sum of individual errors, because
some of them are likely to be positive & some negative.

Error in the results from arithmetic operations ≠ ∑ individual errors

 We expect some cancellation of errors.

6
GIST Anal Chem 9/17/2016

Propagation of uncertainty from random error

: A general rule (Appendix C)
13

 Suppose you wish to calculate the function, F, of several experimental quantities, x, y, z, ….

 If the errors (ex, ey, ez, …) in x, y, z, … are small, random, and independent of one another,
then the uncertainty, eF, in the function F is approximately:

 The quantities in parenthesis are partial derivatives, which are calculated in the same manner as ordinary
derivatives, except that all but one variable are treated as constants.
3 , then ⁄ 3 , ⁄ 3 2 6

Propagation of uncertainty from random error

: Addition & subtraction
14

 Suppose you perform the following arithmetic, in which the experimental uncertainties are
given by e1, e2, and e3 in parentheses.

 For addition & subtraction, the uncertainty in the answer is obtained from the absolute
uncertainties of the individual terms as follows:

0.03 0.02 0.02 0.04 ∴ 3.06 0.04

 Percent relative uncertainty = (0.041/3.06) × 100 %= 1.3 ∴ 3.06 1%

7
GIST Anal Chem 9/17/2016

Propagation of uncertainty from random error

: Multiplication & division
15

 For multiplication & division, first convert all uncertainties into percent relative
uncertainties.
 Then, calculate the error of the result as follows:

 For example,
1.76 0.03 1.89 0.02
5.64
0.59 0.02
 First, convert absolute uncertainties into percent relative uncertainties.
1.76 1. % 1.89 1. %
5.64
0.59 3. %
% 1. 1. 3. 4. % ∴ 5.6 4. %
 Finally, 4. % 5.6 0.04 5.6 0.2 ∴ 5.6 0.2

Propagation of uncertainty from random error

: Mixed operations
16

1.76 0.03 0.59 0.02

0.619 ?
1.89 0.02
 First, work out the difference in the numerator, using absolute uncertainties.
0.03 0.02 0.03
∴ 1.76 0.03 0.59 0.02 1.17 0.03
 Then, covert into percent relative uncertainties.
1.17 0.03 1.17 3. %
0.619 3. % 3. % 1. % 3. %
1.89 0.02 1.89 1. %
 So, the absolute uncertainty is 0.03 0.619 0.02 .

 The final answer can be written as

0.619 0.02 absolute or 0.619 3. % relative
 Because the uncertainty begins in the 0.01 decimal place, it is reasonable to write as
0.62 0.02 absolute or 0.62 3% relative

8
GIST Anal Chem 9/17/2016

Propagation of uncertainty from random error

: The real rule for significant figures
17

 The first digit of the absolute uncertainty is the last significant digit in the answer.

 4 sf in the measurement vs. 3 sf in the answer

0.002 364 0.000 003
0. 094 6 0.0002
0.025 00 0.000 05

 4 sf in the measurement vs. 4 sf in the answer

0.002 664 0.000 003
0. 106 6 0.0002
0.025 00 0.000 05

 3 sf in the measurement vs. 4 sf in the answer

0.821 0.002
1.022 0.004
0.803 0.002

Propagation of uncertainty from random error

: Exponents & logarithms
18

 For the function , the percent relative uncertainty in % is:

→% %
 % 3% → % 2 3% 6%

 In calculations involving log , ln , 10 , and , use relative uncertainty ( ⁄ ).

1
log → 0.434 29
ln 10
ln →

10 → ln 10 2.302 6

9
GIST Anal Chem 9/17/2016

Propagation of uncertainty from random error

: Summary
19

5. Propagation of Uncertainty from Systematic Error

: Uncertainty in atomic mass
20

The atomic mass of oxygen in the periodic table: 15.999 4 ± 0.000 3 g/mol

 The uncertainty is NOT mainly from random error in measuring the atomic mass.

 The uncertainty is predominantly from isotopic variation in samples of oxygen from

different sources.

 That is, oxygen from one source could have a mean atomic mass of 15.999 1 & oxygen
from another source could have an atomic mass of 15.999 7.

 The atomic mass of oxygen in a particular lot of reagent has a systematic uncertainty.

 Oxygen atomic mass from different sources approximates a rectangular distribution.

10
GIST Anal Chem 9/17/2016

Propagation of uncertainty from systematic error

: Rectangular distribution
21

 There is approximately equal probability of finding any atomic mass between 15.999 1 &
15.999 7 and negligible probability outside of this range.

 The mean value (15.999 4), a = The range on either side of the mean (0.000 3)
 Standard (Std.) uncertainty (Std. deviation of uncertainty) for rectangular distribution = ⁄
∴ 0.000 3⁄ 3 0.000 17

Propagation of uncertainty from systematic error

: Uncertainty in molecular mass (same atoms)
22

 Uncertainty in mass of n identical atoms

= n × (std. uncertainty in atomic mass)
= n × (uncertainty listed in periodic table)/ 3

 O2 (15.999 4 ± 0.000 3)
 Std. uncertainty of O: 0.000 17
 Uncertainty of O2:
2 0.000 17 . (NOT 0.000 17 0.000 17 0.00024

 For systematic uncertainty, we add the uncertainties of each term in a sum or difference.

11
GIST Anal Chem 9/17/2016

Propagation of uncertainty from systematic error

: Uncertainty in molecular mass (different atoms)
23

 The standard uncertainty in Mw of C2H4.

 Uncertainty in the periodic table is C: 0.000 8 g/mol & H: 0.000 07 g/mol.
 Std. uncertainty for C & H:
C: 0.000 8/ 3 = 0.000 46, H: 0.000 07/ 3 = 0.000 040
 Uncertainties in the masses of the atoms in C2H4:

 Use the rule for propagation of random uncertainty for the sum of atomic masses of different elements
because uncertainties for different elements are independent.
28.053 16 0.000 92 0.000 16 28.053 16 0.000 93
∴ 28.053 2 0.000 9 g/mol

Propagation of uncertainty from systematic error

: Multiple deliveries from one pipet (Triangular distribution)
24

 A 25-mL Class A volumetric pipet is certified by the manufacturer to deliver:

25.00 ± 0.03 mL

 The volume delivered by a given pipet is reproducible, but can be in the range 24.97 to
25.03 mL.

 However, the pipet manufacturer works hard to make the volume close to 25.00 mL.

 In this case, we approximate the volume of a large number of pipets by the triangular
distribution.

12
GIST Anal Chem 9/17/2016

Propagation of uncertainty from systematic error

: Triangular distribution
25

 The probability falls off approximately in a linear manner as the volume deviates from 25.00 mL
& is negligible outside of 25.00 ± 0.03 mL.

 the mean value (25.00), a = the range on either side of the mean (0.03).
 Standard (Std.) uncertainty (Std. deviation of uncertainty) for triangular distribution = ⁄
∴ 0.03⁄ 6 0.012 mL

Propagation of uncertainty
: Uncalibrated vs. calibrated pipet
26

 The uncertainty in 100 mL when you use  The uncertainty in 100 mL when you use
an uncalibrated 25-mL Class A an calibrated 25-mL Class A volumetric
volumetric pipet (standard uncertainty pipet (24.991 0.006 mL) four times.
= 0.012 mL) four times.
 Calibration improves certainty by removing
 The difference between 25.00 mL & the actual systematic error.
volume delivered by a particular pipet is a  Thus, the uncertainty here is random error.
systematic error.
 The std. uncertainty:
 The std. uncertainty: 0.006 0.006 0.006 0.006
4 0.012 0.048 mL 0.012 mL
∴ 100.00 0.05 mL ∴ 99.964 0.012 mL

13
GIST Anal Chem 9/17/2016

Appendix 3-1: Calculating the Combined Uncertainty

(found at http://www.measurementuncertainty.org/mu/QUAM2000-1.pdf)
27

 Standard uncertainties
 Before combinations, all uncertainty contributions must be expressed as standard uncertainties, that is, as
standard deviations.
 Rectangular distribution
 If limits of ±a are given without a confidence level & there is reason to expect that extreme values are
likely, it is normally appropriate to assume a rectangular distribution, with a standard deviation of a/√3.
 Example: A 10 ml Grade A volumetric flask is certified to within ±0.2 ml. The standard deviation is 0.2/√3
= 0.12 ml.
 Triangular distribution
 If limits of ±a are given without a confidence level, but there is reason to expect that extreme values are
unlikely, it is normally appropriate to assume a triangular distribution, with a standard deviation of a/√6.
 Example: A 10 ml Grade A volumetric flask is certified to within ±0.2 ml, but routine in-house checks
shows that extreme values are rare. The standard deviation is 0.2/√6 = 0.08 ml.

14