E 185

All rights reserved.
May not be reproduced in any form without permission from the publisher, except fair uses permitted under U.S. or applicable copyright law.
Copyright 2003. Royal Society of Chemistry.
EBSCO Publishing : eBook Collection (EBSCOhost) - printed on 4/27/2022 6:57 AM via POLITCNICO DO PORTO
AN: 496138 ; Mullins, Eamonn.; Statistics for the Quality Control Chemistry Laboratory
Account: ns000503.main.ehost
Statistics for the Quality Control Chemistry Laboratory
EBSCOhost - printed on 4/27/2022 6:57 AM via POLITCNICO DO PORTO. All use subject to https://www.ebsco.com/terms-of-use
Statistics for the Quality Control
Chemistry Laboratory
Eamonn Mullins
Trinity College, Dublin, Ireland
ISBN 0-85404-671-2
A catalogue record for this book is available from the British Library
# The Royal Society of Chemistry 2003
All rights reserved
Apart from any fair dealing for the purpose of research or private study, or criticism or
review as permitted under the terms of the UK Copyright, Designs and Patents Act, 1988,
this publication may not be reproduced, stored or transmitted, in any form or by any means,
without the prior permission in writing of The Royal Society of Chemistry, or in the case of
reprographic reproduction only in accordance with the terms of the licences issued by the
Copyright Licensing Agency in the UK, or in accordance with the terms of the licences
issued by the appropriate Reproduction Rights Organization outside the UK. Enquiries
concerning reproduction outside the terms stated here should be sent to The Royal Society
of Chemistry at the address printed on this page.
Published by The Royal Society of Chemistry,
Thomas Graham House, Science Park, Milton Road, Cambridge CB4 0WF, UK
Registered Charity Number 207890
For further information see our web site at www.rsc.org
Typeset by Alden Typeset, Northampton, UK
Printed by TJ International Ltd, Padstow, Cornwall, UK
Preface
This book is intended to be a simple, though not simplistic, introduction

to the statistical methods that are routinely applied in analytical
chemistry Quality Control laboratories. While it is strongly influenced
by my experiences of teaching short courses to laboratory scientists in
pharmaceutical companies, the ideas are quite general and would find
application in virtually every type of analytical chemistry laboratory. I
have included data from different application areas, based on a wide
range of analytical techniques. I hope, therefore, that the book will have
a broad appeal and serve the needs of a large laboratory audience.
The book is oriented towards the needs of analysts working in QC
laboratories rather than towards a wider research-oriented readership.
Accordingly, it focuses on a small number of statistical ideas and
methods and explores their uses in the analytical laboratory. The selected
methods are important aids in method development, method validation
and trouble-shooting. The book strongly emphasises simple graphical
methods of data analysis, such as control charts, which are a key tool in
Internal Laboratory Quality Control and which are a fundamental
requirement in laboratory accreditation. A large part of the book is
concerned with the design and analysis of laboratory experiments. The
coverage ranges from the simplest studies, requiring a single system
parameter change, to robustness/ruggedness studies, involving simul-
taneous changes to many system parameters as an integral part of the
method validation process. The approach taken focuses on the statistical
ideas rather than on the underlying mathematics. Practical case studies
are used throughout to illustrate the ideas in action. A short introduction
to the eight chapters is given below.
Chapter 1 establishes the statistical terminology which is basic to the
description of the quality of analytical results. It introduces the idea of a
statistical model (the Normal distribution) as a basis for thinking about
v
vi Preface
analytical data and describes some of the more commonly used measures
of analytical precision, which are related to this model.
Chapter 2 is mainly about control charts. The chapter focuses entirely
on Shewhart charts, because these are the most widely used and most
easily understood charts. Interested readers will be able to develop their
knowledge of other charting techniques, having established a sound
understanding of the basic material covered here. This approach
characterizes the book as a whole. I have chosen to provide an extended
discussion, including a detailed case study, of the use of one type of chart,
rather than to provide succinct introductions to several different types of
chart. The chapter also contains a short introduction to proficiency tests.
Chapter 3 is an introduction to some of the ideas of statistical
inference. It covers three main areas: statistical tests, confidence intervals
and the determination of sample size. These ideas are fundamental and
are developed in various ways in the remaining chapters.
Chapter 4 is an introduction to the statistical aspects of experimental
design. First, it builds on the methods introduced in Chapter 3 by
showing how statistical tests and the associated confidence intervals may
be used to analyze the data generated by designed experiments – the
studies are of the simplest type, involving a change to a single aspect of
the analytical system. Next it discusses how sample sizes appropriate for
such studies can be determined in advance. It then discusses design
aspects such as randomization, pairing and appropriate measures of
precision. Throughout, there is a focus on validating the assumptions of
the simple statistical models that underlie the tests and confidence
intervals. Residual plots and tests for Normality are key tools in doing so.
Chapter 5 develops the arguments of Chapter 4 to discuss how
complex systems may be investigated using two-level factorial designs.
These designs are important tools for method development, method
validation and trouble-shooting. Full factorial designs for investigating
relatively small numbers of system parameters are discussed in detail.
The fractional factorial designs that form the basis of the designs
commonly used for robustness/ruggedness testing are then discussed.
Chapter 6 is concerned with the use of regression analysis for
modelling relationships between variables. The technique is introduced
in the context of stability testing of pharmaceutical products – a case
study concerned with the establishment of the shelf life of a drug is
described. The application of regression analysis to calibration is then
discussed. The use of residual analysis for validating the statistical
models is emphasized. Examples are shown of how the commonly
encountered problems of non-linearity and changing variability may be
detected. Methods for dealing with these problems are then introduced:
Preface vii
weighted least squares to allow for changing response variability and

multiple regression to model non-linearity. A short introduction to the
fitting of response surfaces is also given.
Chapter 7 extends the discussion of experimental designs. It introduces
(fixed effects) Analysis of Variance (ANOVA), which is used to analyze
multi-level factorial designs – these are a simple extension of the two-
level designs of Chapter 5. In this context, the idea of blocking is
discussed–this is an extension of the discussion of paired comparisons,
introduced in Chapter 4. Nested or hierarchical designs are then
introduced. These designs implicitly underlie the discussion of control
charts in Chapter 2 and are also key to the estimation of the various
measures of precision discussed in Chapter 1.
Chapter 8 discusses quantitative measures of the quality of
measurements produced by a laboratory. The long-standing approach
of estimating the repeatability and reproducibility standard deviations of
a method by conducting a collaborative inter-laboratory trial is
discussed first. This is followed by an introduction to the statistical
ideas used in estimating ‘measurement uncertainty’. The use of
collaborative trial data in the estimation of measurement uncertainty is
then discussed. The final section reviews many of the ideas discussed
earlier in the book using the framework of a hierarchical statistical model
of the measurement process. Sophisticated readers may find this section
a useful introductory overview of some of the important issues discussed
in the book. Novices would probably find it abstract and unhelpful.
Careful study of the rest of the book will, I hope, change this.
I have assumed that virtually everyone working in a technical
environment will have access to computers. All the calculations required
for this book may be implemented in a spreadsheet, though I would
recommend, in preference, use of a validated statistics package.
Accordingly, while formulae suitable for hand calculations are
presented, the book assumes that calculations will be carried out using
a computer, and focuses on the interpretation of the results. The data
analyses in the book were carried out using Minitab, but any statistical
package might be used for the purpose – the book is not intended to
teach Minitab, per se.
The book is not a comprehensive account of all the statistical methods
that are likely to be of value in the analytical laboratory. If it is
successful, it will give the reader a strong grasp of the concept of
statistical variation in laboratory data and of the value of simple
statistical ideas and methods in thinking about and manipulating such
data. I have deliberately limited the range to include only those topics
that I have encountered in use in the laboratories with which I have had
viii Preface
dealings. Of course, this means that experienced readers will find that
topics they consider important are absent, but a line had to be drawn and
this seemed a reasonable decision criterion. For some, the omission of
any discussion of multivariate statistical methods (chemometrics) will be
notable. However, a good understanding of the simple ideas discussed in
this book is, in my opinion, a minimum requirement before the more
sophisticated methods are likely to be used confidently or correctly.
Accordingly, I have preferred to focus the discussion on the simpler
ideas.
Acknowledgements
My principal debt is obviously to those who developed the statistical

ideas discussed in the book. Next, I am indebted to those who taught me
these ideas – both in the classroom and through their writing. Of these,
the first group is composed of colleagues and former colleagues in Trinity
College, Dublin. In particular, I want to single out Professor Gordon
Foster, who set up the Department of Statistics, and from whom I
learned so much over many years of interaction. I would find it difficult
to produce a list of the statistical writing that has most influenced me,
though some books are given in the references. However, the writings of
Professor George Box would undoubtedly be at the top of any such list.
The references in the text list papers from which I learned much about
the application of statistical methods to analytical chemistry data. I am
conscious of how often the statistical sub-committee of the Analytical
Methods Committee of the RSC is listed and, in particular, of the
number of times I have made reference to the work of its Chairman,
Professor Michael Thompson and his co-workers. Readers will find it
profitable to follow-up these references.
This book arises directly out of my teaching in-house short courses in
industry. These have covered a wide range of industries and statistical
topics and have involved teaching both manufacturing engineers and
chemists, and laboratory scientists. All the courses have helped to shape
the current book, since the statistical ideas are essentially the same
whether the objective is to evaluate, monitor, trouble-shoot or optimize
either a manufacturing or an analytical system. Accordingly, I am
indebted first to those who commissioned the courses and then to the
attendees. I doubt if I have ever taught a short course without learning
something from the participants. It would be impossible though to thank
by name all those who have influenced me in this way. However, some
people who have very directly influenced the book must be thanked.
ix
x Acknowledgements
I am indebted to the following friends, clients and colleagues who have

either commissioned courses, provided me with data, or who have
clarified analytical chemistry ideas for me. To some I am heavily
indebted, but I am grateful to all of them for what they have given me:
Jill Ahearne, Norman Allott, Norah Blount, John Bohan, John Buckley,
Dorothy Claffey, Margaret Connolly, Marion Cullinane, Martin
Danaher, Tom Dempsey, Robert Dunne, Mary English, Lynn Feery,
Marion Finn, Jacintha Griffin, Rosemary Hayden, Denise Heneghan,
Arlene Hynes, Fintan Keegan, Jim Kelly, Kevin Kinnane, Des McAteer,
Ger McCann, Ken McCartney, Kevin McNamara, Michael Metzler,
Liam Murphy, John O’Connor, Michael O’Dea, Tom O’Hara, Michael
O’Keeffe, Marie O’Rourke, John Prendergast, Joe Rowley, Colette
Ryan, Eva Ryan, Tom Ryan, Kevin Shelly and Sarah Tait. Andrew
Mullins and Paul McNicholas provided research assistance for which I
am grateful. I remember with pleasure many conversations on topics
relevant to the book with my former MSc student Joe Vale. My colleague
Myra O’Regan, by taking on the headship of the Department of
Statistics in Trinity College for six months, allowed me to take sabbatical
leave in order to concentrate on finishing the book – I am grateful for her
generosity in doing so.
I would like to thank Janet Freshwater, Katrina Turner and the
editorial and production staff of the Royal Society of Chemistry for
turning what was undoubtedly a difficult typescript into the finished
book you are now holding. The book is heavily dependent on data
extracted from two analytical chemistry journals, The Analyst and The
Journal of the Association of Official Analytical Chemists – I am grateful
to the editors for permission to reproduce the data extracted from the
articles cited in the references. I am grateful to Martin Danaher and
Michael O’Keeffe of the Irish Food Centre for providing me with a
copy of Figure 6.29 and to Dr A. Lamberty of the Institute for Reference
Materials and Measurements, Belgium, for providing a copy of
Figure 1.2. I am grateful to Pearson Education for permission to
reproduce Figure 3.15 from the book Statistics for the Life Sciences by
Professor Myra L. Samuels.
Two friends deserve special thanks. Jack Doyle of the Irish State
Laboratory has been a constant source of information on matters related
to analytical chemistry, over many years. He also provided valuable
comments on the first draft of Chapter 8. Michael Stuart, as a teacher,
colleague and friend has had more influence than anyone else on my
thinking about statistics. He read earlier drafts of all the chapters and
provided detailed criticisms and helpful suggestions. It is certainly a better
book than it would otherwise have been without his many contributions.
Acknowledgements xi
Finally, I must mention my family. First, my sisters for their friendship

and support, as Tom Lehrer might say, from childbirth! My brother-in-
law Tom Mayhew was a more than generous host when, as a student, I
needed to earn enough to survive the next academic year. For this and all
his kindness to my family I will always be grateful. My immediate family,
Andrew, Deirdre and Genevieve, have had to live with the writing and
re-writing of this book for too long, but finally, here it is, and it is
dedicated to them.
Eamonn Mullins
Contents
Chapter 1 Variability in Analytical Measurements 1
1.1 Introduction 1
1.2 An Example of Measurement Variability 3
1.3 Describing Measurement Error 4
1.3.1 A Schematic Inter-laboratory Study 6
1.4 Sources of Analytical Variability 8
1.5 Measuring Precision 10
1.5.1 The Standard Deviation as a Measure
of Precision 13
1.5.2 Variation of Precision with Concentration 15
1.5.3 Measures of Repeatability and
Reproducibility 19
1.6 Case Study: Estimating Repeatability from
Historical Data 22
1.7 Improving Precision by Replication 26
1.8 Conclusion 29
1.9 Review Exercises 29
1.10 References 33
Chapter 2 Control Charts in the Analytical Laboratory 35

2.1 Introduction 35
2.2 Examples of Control Charts 36
2.3 The Theory Underlying the Control Limits 39
2.4 Setting Up Control Charts 42
2.4.1 Calculating the Limits 42
2.4.2 Data Scrutiny 45
2.4.3 Sample Size 47
2.5 Monitoring Precision 50
2.5.1 Range Charts 50
2.5.2 The Nature of Replicates 51
2.5.3 Standard Deviation Charts 52
2.6 Case Study 55
2.7 Control Chart Performance 61
xiii
xiv Contents
2.7.1 Average Run Length Analysis 61
2.7.2 How Many Control Samples? 64
2.8 Learning from Control Charts 66
2.8.1 Improving Precision by Replication:
Revisited 66
2.8.2 Obtaining Measures of Precision from
Control Charts 72
2.8.3 Using Control Charts 74
2.8.4 Concluding Remarks 76
2.9 Proficiency Testing 76
2.9.1 Overview 76
2.9.2 Technical Issues 78
2.9.3 Concluding Remarks 79
2.10 Conclusion 80
2.12 References 84
Chapter 3 Some Important Statistical Ideas 87

3.1 Introduction 87
3.2 Statistical Significance Tests 88
3.2.1 Example 1: A Method Validation Study 88
3.2.2 Example 2: Acceptance Sampling 93
3.2.3 Summary 95
3.3 Determining Sample Size 96
3.3.1 The Nature of the Problem 96
3.3.2 Using the Sample Size Table 98
3.3.3 Discussion 100
3.3.4 Some Useful Graphs: Power Curves 101
3.4 Confidence Intervals for Means 103
3.4.1 Example 1: Estimating the Average
Potency of a Pharmaceutical Material 104
3.4.2 Example 2: The Method Validation
Study Revisited–Estimating Bias 108
3.4.3 Example 3: Estimating the Potency
of a Pharmaceutical Material: Revisited 110
3.4.4 Example 4: Error Bounds for
Routine Test Results 112
3.5 Sampling 114
3.6 Confidence Intervals for Standard Deviations 116
3.7 Checking Normality 121
3.7.1 Normal Probability Plots 122
3.7.2 A Significance Test for Normality 125
3.7.3 Departures from Normality 126
3.7.4 Transformations 128
Contents xv
3.8 Concluding Remarks 130
3.10 References 132
Chapter 4 Simple Comparative Studies 135

4.1 Introduction 135
4.2 A Typical Comparative Study 135
4.2.1 A Statistical Significance Test for
Comparing Method Means 137
4.2.2 Estimating the Difference in
Recovery Rates 140
4.2.3 Comparing Standard Deviations 143
4.2.4 Comparing Means when Standard
Deviations are Unequal 145
4.2.5 Validating the Assumption of Normality 147
4.3 Paired Comparisons 148
4.3.1 A Trouble-shooting Exercise 149
4.3.2 Case Study 154
4.4 Sample Size for Comparative Studies 159
4.4.1 Comparing Means of Two
Independent Groups 159
4.4.2 Paired Studies of Relative Bias 164
4.4.3 Sample sizes for Comparing
Standard Deviations 166
4.5 Some Comments on Study Design 168
4.5.1 Experimental Run Order 169
4.5.2 Appropriate Measures of Precision 171
4.5.3 Representativeness 179
4.8 References 183
Chapter 5 Studying Complex Systems 185

5.2 Statistical or Traditional Designs? 186
5.3 The 22 Design 189
5.3.1 An Example 190
5.3.2 Model Validation 197
5.3.3 Organizing the Calculations 198
5.4 The 23 Design 201
5.4.1 An Example 201
5.4.2 Data Analysis 207
5.4.3 Model Validation 210
5.5 Sample Size for Factorial Designs 212
xvi Contents
5.6 Experiments with many Factors 215
5.7 Fractional Factorial Designs 219
5.7.1 A Simple Example 219
5.7.2 The 25-1 Design 224
5.7.3 Blocking 228
5.8 Ruggedness Testing 230
5.8.1 Designing Ruggedness Tests 231
5.8.2 Example 1 233
5.8.3 Example 2 235
5.11 References 246
Chapter 6 Fitting Equations to Data 247

6.2 Regression Analysis 248
6.2.1 Introductory Example 248
6.2.2 Using the Regression Line 253
6.2.3 Analysis of Variance 260
6.3 Calibration 264
6.3.1 Example 265
6.3.2 Error Bounds for the Estimated
Concentration 267
6.3.3 Zero-intercept Calibration Lines 270
6.4 Detection Limit 275
6.5 Residual Analysis 278
6.6 Weighted Regression 286
6.6.1 Fitting a Calibration Line by WLS 287
6.6.2 Is Weighting Worthwhile? 290
6.7 Non-linear Relationships 295
6.7.1 A Single Predictor Variable 296
6.7.2 A ‘Lack-of-fit’ Test 299
6.7.3 Response Surface Modelling 301
6.10 References 307
Chapter 7 The Design and Analysis of Laboratory

Studies Re-visited 309
7.2 Comparing Several Means 310
7.2.1 Example 1: A Laboratory
Comparison Study 310
Contents xvii
7.2.2 Multiple Comparisons 320
7.2.3 Example 2: A Method
Development Study 321
7.3 Multi-factor Studies 324
7.3.1 Example 1: The GC Study Re-visited 324
7.3.2 Example 2: A 3 · 3 Study 330
7.3.3 Example 3: A 23 Study 334
7.4 Blocking in Experimental Design 338
7.4.1 Example 1: The GC Development
Study Re-visited, Again! 338
7.4.2 Example 2: Paired t-Tests Revisited 342
7.5 Estimating Components of Test
Result Variability 343
7.5.1 Example 1: Control Charts 344
7.5.2 Example 2: Three Variance Components 351
7.6 Conclusion 360
7.8 References 364
Chapter 8 Assessing Measurement Quality 365

8.2 Inter-laboratory Collaborative Trials 366
8.2.1 Estimating the Reproducibility
Standard Deviation 366
8.2.2 Data Scrutiny 369
8.2.3 Measuring the Trueness of a Method 375
8.3 Measurement Uncertainty 377
8.3.1 Example: Preparing a Stock Solution 379
8.3.2 Discussion 384
8.4 An Integrated Approach to
Measurement Uncertainty 386
8.7 References 396
Solutions to Exercises 397
Appendix: Statistical Tables 431
Subject Index 447

E 185

Uploaded by

Copyright:

Available Formats

E 185

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

E 185

Uploaded by

Copyright:

Available Formats

All rights reserved.

This book is intended to be a simple, though not simplistic, introduction

weighted least squares to allow for changing response variability and

My principal debt is obviously to those who developed the statistical

I am indebted to the following friends, clients and colleagues who have

Finally, I must mention my family. First, my sisters for their friendship

Chapter 2 Control Charts in the Analytical Laboratory 35

Chapter 3 Some Important Statistical Ideas 87

Chapter 4 Simple Comparative Studies 135

Chapter 5 Studying Complex Systems 185

Chapter 6 Fitting Equations to Data 247

Chapter 7 The Design and Analysis of Laboratory

Chapter 8 Assessing Measurement Quality 365

Solutions to Exercises 397

Appendix: Statistical Tables 431

Subject Index 447

You might also like