CAN THO UNIVERSITY
COLLGE OF AQUACULTURE AND FISHERIES
Lecture 9: RELIABILITY OF MEASUREMENTS
AND QUALITY CONTROL
WATER QUALITY ANALYSIS
(for Advance Aquaculture Program)
1
Concepts
• Quality assurance: the steps taken during an analysis
to ensure that the analysis is under control and that it is
properly monitored.
• Quality control: Those steps taken to ensure that an
analysis is under statistical control.
• Good laboratory practices: Those general laboratory
procedures that, when followed, help ensure the quality
of analytical work.
Reliability of measurements
• Determinate errors (systematic errors) are always in
the same direction and of about same magnitude.
Balance errors
Glassware errors
Personal errors
• The magnitudes of determinate errors can ussually be
ascertained and appropriate corections applied.
Reliability of measurements
• Indeterminate errors (accidental errors) are random
errors which occur in all analyses.
• Indeterminate errors cannot be isolated and corected.
• Indeterminate errors may be treated statistically to
ascertain if the magnitude of the error will affect the
reliability of analytical data.
Reliability of measurements
• Average deviation
d1 d 2 ... d n
d av
n
where:
dav: Average deviation
d : The absolute of the deviation of individual
determination from mean
n : The number of determination
Reliability of measurements
• Average deviation
Example: DO concentration in water was measured
respectively as 8.01, 8.02, 8.01, 8.05 and 8.03 mg/L
8.01 8.02 8.01 8.05 8.03
x 8.02 mg/L
5
0.01 0.00 0.01 0.03 0.01
d av 0.01 mg/L
5
• It is proper to write the mean as 8.02 mg/L because the
deviation digits only in the second decimal place
Reliability of measurements
• Standard deviation Where
Large number of observations , s: Standard deviation
(x 1 μ)2 (x 2 μ)2 ... (x n μ)2 , x: the mean of
observations
n x1, x2…xn: individual
Small number of observations observation
n: number of
(x 1 x ) 2 (x 2 x ) 2 ... (x n x ) 2 observations
s
n 1 n-1: degree of freedom
Reliability of measurements
• Standard deviation Example
A more practical formula Replicate calcium determination gave 21.1,
for making calculations is 20.8,20.9,21.4, and 21.2 mg/L
105.4
x
2
x 21.1 mg/L
x n
2
5
s
n 1 2,222.06 2,221.83
s 0.24 mg/L
4
• The smaller the deviation, the more precise of determinations
Reliability of measurements
• Confidence intervals Example
C.I.x 0.24 x 2.776 0.67
C.I.x s.t α
Confidence intervals for individual determination is
21.10.67 at 95% confidence level.
C.I.x s x .t α
0.24
sx 0.11
• Standard error of mean 5
s C.I.x 0.11 x 2.776 0.31
sx Confidence intervals for mean is 21.10.31 at
n 95% confidence level.
Reliability of measurements
• Coefficient of variance (relative standard deviation)
s
CV 100
x
CV permits easy coparisons of precision among groups
of replicate determination of different samples which
have different mean
Reliability of measurements
• Example: Phosphorus determinations on three water samples gave
the following mean and standard deviation.
0.15
Sample 1 : 1.02 0.15 mg/L CV 100 15%
1.02
0.12
Sample 1 : 0.53 0.12 mg/L CV 100 23%
0.53
0.04
Sample 3 : 0.11 0.04 mg/L CV 100 36%
0.11
The higher the concentration of analyte, the higher the precision
The lower the coefficient of variation, the higher the precision
Quality Control
Quality control by precision and accuracy checks
Precision check:
• Prepare standard solutions of substance to be measured from
lab. reagents and distilled water.
• Estimate the precision on three samples: one low, one
intermediate and one high in concentration of substance to be
measured.
• Make a number of repetitive measurements (5-10 is aceptable)
on sample to calculate the mean, standard deviation or
confidence interval for individual measurements.
Quality control by precision and accuracy checks
Replicate Total alkalinity (mg/L)
Sample A Sample B Sample C
1 18.0 64.9 155.6
2 17.8 64.4 155.4
3 17.8 64.5 156.2
4 18.0 64.1 156.8
5 18.0 64.1 156.8
6 17.8 64.5 156.6
7 17.8 64.4 156.4
x 17.9 64.4 156.1
s 0.11 0.27 0.53
95% C.I 0.27 0.66 1.30
C.V 0.61% 0.42% 0.34%
Quality control by precision and accuracy checks
Precision check:
• High concentration can be analyzed with slightly beter
ptecision than water with low cocentration.
• In the range from 17.9 to 156.1 mg/L total alkalinity, a
measured value should fall within 1.30 mg/L 0f mean
95% of the time
Quality control by precision and accuracy checks
Accuracy check:
• The accuracy of a procedure can be checked by adding a
known amount of the substrance to be measured to distilled
water. Analyze and determine how close the measured value
approaches the true value.
• Accuracy of a procedure can be achieved by determining the
concentratrion of the substance in natural water then adding a
known amount of the substance to the natural water and
determining the percentage recovery.
Quality control by precision and accuracy checks
Example: Replicate Calcium (mg/L)
• A water has a measured calcium 1 25.0
concentration of 15.1 mg/L. 2 24.9
• To this water, add 10.0 mg/L calcium 3 24.8
so that the water now contains 25.1 4 25.0
mg/L calcium. 5 24.7
• Replicate determinations are made 6 25.2
and the data are treated as following 7 24.9
24.93 x 24.9
Recovery 100 99.3%
15.1 10
Quality control by precision and accuracy checks
• Periodic checks of precision and accuracy should be made.
• Five to ten percent of samples should be analyzed in duplicate.
• If the duplicate measurements do not agree with known
precision of the method, the results are not reliable and the
problem in analyst's technique must be located and corrected.
• For colorimetric methods, calibration graphs should be verified
frequently by analyzing known concentrations of the substance
and determining whether the original absorbance is obtained.
Quality control by precision and accuracy checks
• Making duplicate or triplicate analyses of all samples is
essentially worthless. Duplicate analyses provide no estimate
of accuracy.
Shewart quality control charts
• Quality control chart for precision:
Step 1st: Make duplicate analyses of water samples
during normal laboratory operation. The data should be
collected over several weeks until at least 20 to 25 test
results are available. (Example: TAN determination)
Step 2nd: Calculate the average range (see table in next
slide)
R
R R
1.29
0.05 mg/L
n 25
Shewart quality control charts
Date Result 1 Result 2 Range Date Result 1 Result 2 Range
July 2 0.51 0.47 0.04 July 17 0.62 0.58 0.04
July 3 0.25 0.20 0.05 July 19 1.00 0.92 0.08
July 3 0.11 0.09 0.02 July 23 0.78 0.71 0.07
July 6 1.25 0.92 0.13 July 23 0.98 0.92 0.06
July 9 0.82 0.95 0.13 July 24 0.68 0.72 0.04
July 9 0.75 0.74 0.01 July 26 1.25 1.31 0.06
July 10 0.44 0.44 0.00 July 27 0.05 0.05 0.00
July 11 0.36 0.38 0.02 July 31 1.33 1.25 0.08
July 12 2.13 2.05 0.08 August 2 1.62 1.74 0.12
July 12 1.50 1.55 0.05 August 3 0.45 0.42 0.03
July 13 0.09 0.06 0.03 August 3 0.63 0.66 0.04
July 16 0.35 0.37 0.02 August 7 0.80 0.75 0.05
July 17 0.50 0.54 0.04
Shewart quality control charts
Step 3rd: Calculate the upper control limit on the range
UCLR 3.27 R UCLR (3.27)(0.052) 0.17 mg/L
Step 4th: Calculate the upper warning limit on the range
UWLR 0.67(3.27 R R ) R
or UWL R 2.52 R UWLR 2.52 (0.052) 0.13 mg/L
Step 5th: Graph R, UWLR, UCLR (see the graph in next
slide)
Shewart quality control charts
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Shewart quality control charts
• To use the chart, the analyst continues to make duplicate
analyses TAN on about 10 percent of samples an the results
are plotted over time (see the table and the chart below).
Date Date
August 1 0.57 0.63 0.06 August 9 0.75 0.69 0.06
August 1 1.11 1.04 0.07 August 13 0.60 0.51 0.09
August 2 0.15 0.15 0.00 August 14 0.15 0.18 0.03
August 3 1.05 1.12 0.07 August 17 0.50 0.43 0.07
August 6 0.33 0.29 0.04 August 20 0.83 0.74 0.08
August 7 1.25 1.27 0.02 August 21 0.17 0.19 0.02
August 7 0.13 0.09 0.04
Shewart quality control charts
Points on graph represent plot of data when quality
control is being maintained
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Shewart quality control charts
Points on graph represent plot of data when quality
control is not being maintained.
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Shewart quality control charts
• Quality control chart for accuracy:
Quality control chart for accuracy are similar to those
for precision. The analyst may spike about 10 percent
of samples and obtain the difference between expected
and measured values.
Shewart quality control charts
Example:
Sample contains 0.50 mg/L of TAN.
After spiking with an additional 0.20 mg/L of TAN, a
concentration 0.65 mg/L was measured
Expected value would be the initial value (0.50 mg/L) plus the
concentration of the spike (0.20 mg/L) or 0.70 mg/L. The range
would be the expected value (0.70) minus measued value
(0.65), or 0.05 mg/L.
The chart is then prepared and used in the same manner
described above for precision.
Shewart quality control charts
An alternative method for assessing quality control for
accuracy is to periodically prepare standard solutions
and calculate R as the difference between the standard
concentration and the analytical value obtained from
the standard solution.
However, this method is not as good as the technique
of spiking water sample.
Control Chart Interpretation
1. The method is in control
• The control value is within the UWL
• The control value is between UWL and UCL and
two previous control value were within UWL.
Control Chart Interpretation
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Control Chart Interpretation
0.20
UCLR 0.17
0.15
UWLR 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Control Chart Interpretation
2. The method is in control but can be regarded as out of
statistical control.
• Seven control values in consecutive order gradually
increase or decrease.
• Ten of 11 consecutive control values are lying on the
same side of central line
Control Chart Interpretation
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Control Chart Interpretation
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Control Chart Interpretation
3. Method is out of control
• The control value is outside of UCL
• The control value is between the UWL and UCL and at
least one of two previous control values is also
between UWL and UCL (the rule two out of three)
Control Chart Interpretation
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
Control Chart Interpretation
0.20
UCL R 0.17
0.15
UWL R 0.13
0.10
0.05 R 0.05
0
1 5 10 15 20 25
