WELCOME
INTRODUCTION TO STATISTICAL VALIDATION
JIJO PAUL K
APPROACHES TO VALIDATION
Statistical or Retrospective Validation – Based on
historical data .
Experimental Validation – Based on plant trials data
STATISTICAL VALIDATION
To be done before carrying out experimental validation.
Retrospective validation is an approach based on analysis of historical data.
More the no of trials better will be the statistical validation results.
The results will indicate that whether the process is under control or not.
STATISTICAL TOOLS CONSIDERED FOR VALIDATION
Control charts
Capability study
Scatter diagrams
STATISTICAL TERMS : Standard Deviation – Sigma ( s ) Mean - µ Slope - b Y Intercept - a Upper Standard Limit - USL Lower Standard Limit - LSL Process capability index - Cp Process performance index upper - Cp K upper
Process performance index lower - Cp K lower
Correlation Coefficient – r Coefficient of determination - R2
Regression line y = a + bx Control Ratio - CR
+3
-3Time
CONTROL CHARTS - APPLICATIONS
“Early Warning”
Assures that Process is Working
Provides Information on “Process Capability”
Distinguishes between common and spl cause problems
CAPABILITY STUDYCAPABILITY STUDY
Capability studies are performed to evaluate the ability of a process to consistently meet a specification.
Cp = (allowable range)/6s = (USL - LSL)/6s. Where Cp is the capability index.
Cpk = min[ (USL - m)/3s, (m - LSL)/3s ].Where Cpk is the process performance index.
CR = (UCL-LCL)/(USL-LSL). Where CR is control ratio.
+3-3
-3 +3
Good
Poor
CPK>1
CPK<1
PROCESS CAPABILITYPROCESS CAPABILITY
Cp > 1
Cp < 1
CR < 1
CR > 1
LSL USL
LSL USL
EXAMPLES for Control Charts
TRIAL DATA For Control chart 1:
Trial NoRM active ingredient
%Mean ( X - M ) ( X - M )2
1 8.63 8.539 0.091 0.008
2 8.52 8.539 -0.019 0.000
3 8.76 8.539 0.221 0.049
4 8.35 8.539 -0.189 0.036
5 8.45 8.539 -0.089 0.008
6 8.81 8.539 0.271 0.073
7 8.23 8.539 -0.309 0.095
8 8.65 8.539 0.111 0.012
9 8.46 8.539 -0.079 0.006
10 8.53 8.539 -0.009 0.000
Mean 8.539
Std deviation
0.18
.+1s 8.72
.-1s 8.36
.+2s 8.90
.-2s 8.18
.+3s 9.08
.-3s 8.00
LSL 7.80
USL 9.00
Cp 1.12
Cpk1 1.38
Cpk2 0.86
CR 0.90
Mean+/- 1S+/- 2S+/- 3S
TRIAL DATA for control chart 2 :
Trial No Efficiency Mean ( X - M ) ( X - M )2
1 72.3 70.26 2.04 4.16
2 70.5 70.26 0.24 0.06
3 68.6 70.26 -1.66 2.76
4 73.2 70.26 2.94 8.64
5 69.5 70.26 -0.76 0.58
6 72.8 70.26 2.54 6.45
7 67.9 70.26 -2.36 5.57
8 69.3 70.26 -0.96 0.92
9 70.5 70.26 0.24 0.06
10 68.9 70.26 -1.36 1.85
mean 70.26
std dev 1.86
.+1s 72.12
.-1s 68.40
.+2s 73.97
.-2s 66.55
.+3s 75.83
.-3s 64.69
LSL 63.000
USL 75.000
Cp 1.077
Cpk1 1.303
Cpk2 0.851
CR 0.929
Mean+/- 1S+/- 2S+/- 3S
SCATTER DIAGRAMSSCATTER DIAGRAMS
Scatter Diagrams are used to study and identify the possible relationship between the changes observed in two different sets of variables.
SCATTER DIAGRAMS
The coefficient of determination ranges from 0 to 1.
An R2 of 0 means that the dependent variable cannot be predicted from the independent variable.
An R2 of 1 means the dependent variable can be predicted without error from the independent variable.
The quantity r, called the linear correlation coefficient, measures the strength and the direction of a linear relationship between two variables.
Importance of correlation coefficient The value of r is such that -1 < r < +1. The + and – signs are used
for positive linear correlations and negative linear correlations respectively
A correlation greater than 0.8 is generally described as strong,
whereas a correlation less than 0.5 is generally described as weak. The coefficient of determination, r 2, is useful because it gives the
proportion of the variance (fluctuation) of one variable that is predictable from the other variable.
EXAMPLE for Scatter Diagramme
TRIAL DATA FOR SCATTER CHART
EFFICIENCY Vs RM Active ingredient %
Trial NoEfficiency
( X)
RM active
ingredient % (Y)
XY X2 Y2
1 73.5 8.63 634.3 5402.3 74.5
2 71.8 8.52 611.7 5155.2 72.6
3 74 8.76 648.2 5476.0 76.7
4 66 8.35 551.1 4356.0 69.7
5 68.6 8.45 579.7 4706.0 71.4
6 74.2 8.81 653.7 5505.6 77.6
7 64 8.23 526.7 4096.0 67.7
8 71 8.65 614.2 5041.0 74.8
9 71 8.46 600.7 5041.0 71.6
10 72.4 8.53 617.6 5241.8 72.8
Σ 706.5 85.39 6037.9 50020.9 729.4
RESULTS
R 0.910
R2 0.829
Slope (b) 0.047
Y intercept (a) 5.192
Equation y = 5.192 + 0.047x
EXPERIMENTAL VALIDATION
To be done once statistical validation is completed and found to be satisfactory.
Run full process according to SOP n times.
Record all required data in the batch process records (BPR).
Deviations to the procedures must be recorded on the data record forms.
Relevant samples to be given (as per sampling plan) to the QC and results to be recorded in the BPR.
Perform the routine tests associated with the process according to the SOP. Test results must be approved by QC.
EXPERIMENTAL VALIDATION
Attach all data record forms and charts.
Perform all necessary calculations and statistical analysis (pre-determined).
Compare to acceptance criteria.
Prepare deviation report – Justification if any on acceptance and impact of process.
Prepare a process validation report including all relevant data.
The Process must meet all specifications for three consecutive runs. If failed, then validation has to be repeated.
Submit the Document to QA for review and approval.
Thank you