Get Design and Analysis of Experiments Douglas C. Montgomery free all chapters
Get Design and Analysis of Experiments Douglas C. Montgomery free all chapters
Get Design and Analysis of Experiments Douglas C. Montgomery free all chapters
com
https://ebookmeta.com/product/design-and-analysis-of-
experiments-douglas-c-montgomery/
OR CLICK BUTTON
DOWNLOAD NOW
https://ebookmeta.com/product/statistical-methods-in-biology-design-
and-analysis-of-experiments-and-regression-1st-edition-welham/
ebookmeta.com
https://ebookmeta.com/product/perception-and-the-representative-
design-of-psychological-experiments-egon-brunswik/
ebookmeta.com
https://ebookmeta.com/product/the-statistical-analysis-of-quasi-
experiments-christopher-h-achen/
ebookmeta.com
https://ebookmeta.com/product/public-speaking-for-authors-creatives-
and-other-introverts-2nd-edition-joanna-penn/
ebookmeta.com
Pumpkin Julie Murphy
https://ebookmeta.com/product/pumpkin-julie-murphy/
ebookmeta.com
https://ebookmeta.com/product/one-night-with-a-diver-love-on-the-
sunshine-coast-1st-edition-eve-london/
ebookmeta.com
https://ebookmeta.com/product/renewable-energy-1st-edition-rahman/
ebookmeta.com
https://ebookmeta.com/product/lifeboat-1st-edition-micheline-lee/
ebookmeta.com
https://ebookmeta.com/product/hybrid-energy-systems-for-offshore-
applications-1st-edition-dincer/
ebookmeta.com
Dreams Hallucinations Dragons the Unconscious and
Ekphrasis in German Romanticism Ludwig Tieck s Skillful
Study of the Mind Studies on Themes and Motifs in
Literature Joseph D. Rockelmann
https://ebookmeta.com/product/dreams-hallucinations-dragons-the-
unconscious-and-ekphrasis-in-german-romanticism-ludwig-tieck-s-
skillful-study-of-the-mind-studies-on-themes-and-motifs-in-literature-
joseph-d-rockelmann/
ebookmeta.com
Design and Analysis
of Experiments
Ninth Edition
DOUGLAS C. MONTGOMERY
Arizona State University
k
k
k
k
Preface
Audience
k k
This is an introductory textbook dealing with the design and analysis of experiments. It is based on college-level
courses in design of experiments that I have taught for over 40 years at Arizona State University, the University of
Washington, and the Georgia Institute of Technology. It also reflects the methods that I have found useful in my own
professional practice as an engineering and statistical consultant in many areas of science and engineering, including
the research and development activities required for successful technology commercialization and product realization.
The book is intended for students who have completed a first course in statistical methods. This background
course should include at least some techniques of descriptive statistics, the standard sampling distributions, and an
introduction to basic concepts of confidence intervals and hypothesis testing for means and variances. Chapters 10, 11,
and 12 require some familiarity with matrix algebra.
Because the prerequisites are relatively modest, this book can be used in a second course on statistics focusing
on statistical design of experiments for undergraduate students in engineering, the physical and chemical sciences,
statistics, mathematics, and other fields of science. For many years I have taught a course from the book at the first-year
graduate level in engineering. Students in this course come from all of the fields of engineering, materials science,
physics, chemistry, mathematics, operations research life sciences, and statistics. I have also used this book as the
basis of an industrial short course on design of experiments for practicing technical professionals with a wide variety
of backgrounds. There are numerous examples illustrating all of the design and analysis techniques. These examples
are based on real-world applications of experimental design and are drawn from many different fields of engineering
and the sciences. This adds a strong applications flavor to an academic course for engineers and scientists and makes
the book useful as a reference tool for experimenters in a variety of disciplines.
k
k
iv Preface
Empirical Model
I have continued to focus on the connection between the experiment and the model that the experimenter can develop
from the results of the experiment. Engineers (and physical, chemical and life scientists to a large extent) learn about
physical mechanisms and their underlying mechanistic models early in their academic training, and throughout much
of their professional careers they are involved with manipulation of these models. Statistically designed experiments
offer the engineer a valid basis for developing an empirical model of the system being investigated. This empirical
model can then be manipulated (perhaps through a response surface or contour plot, or perhaps mathematically) just
as any other engineering model. I have discovered through many years of teaching that this viewpoint is very effective
in creating enthusiasm in the engineering community for statistically designed experiments. Therefore, the notion of
an underlying empirical model for the experiment and response surfaces appears early in the book and continues to
receive emphasis.
k k
Factorial Designs
I have expanded the material on factorial and fractional factorial designs (Chapters 5–9) in an effort to make the
material flow more effectively from both the reader’s and the instructor’s viewpoint and to place more emphasis on
the empirical model. There is new material on a number of important topics, including follow-up experimentation
following a fractional factorial, nonregular and nonorthogonal designs, and small, efficient resolution IV and V designs.
Nonregular fractions as alternatives to traditional minimum aberration fractions in 16 runs and analysis methods for
these design are discussed and illustrated.
Experimental Design
Throughout the book I have stressed the importance of experimental design as a tool for engineers and scientists to use
for product design and development as well as process development and improvement. The use of experimental design
k
k
Preface v
in developing products that are robust to environmental factors and other sources of variability is illustrated. I believe
that the use of experimental design early in the product cycle can substantially reduce development lead time and cost,
leading to processes and products that perform better in the field and have higher reliability than those developed using
other approaches.
The book contains more material than can be covered comfortably in one course, and I hope that instructors will
be able to either vary the content of each course offering or discuss some topics in greater depth, depending on class
interest. There are problem sets at the end of each chapter. These problems vary in scope from computational exercises,
designed to reinforce the fundamentals, to extensions or elaboration of basic principles.
Course Suggestions
My own course focuses extensively on factorial and fractional factorial designs. Consequently, I usually cover Chapter
1, Chapter 2 (very quickly), most of Chapter 3, Chapter 4 (excluding the material on incomplete blocks and only
mentioning Latin squares briefly), and I discuss Chapters 5 through 8 on factorials and two-level factorial and fractional
factorial designs in detail. To conclude the course, I introduce response surface methodology (Chapter 11) and give
an overview of random effects models (Chapter 13) and nested and split-plot designs (Chapter 14). I always require
the students to complete a term project that involves designing, conducting, and presenting the results of a statistically
designed experiment. I require them to do this in teams because this is the way that much industrial experimentation
is conducted. They must present the results of this project, both orally and in written form.
For this edition I have provided supplemental text material for each chapter of the book. Often, this supplemental
material elaborates on topics that could not be discussed in greater detail in the book. I have also presented some
subjects that do not appear directly in the book, but an introduction to them could prove useful to some students and
professional practitioners. Some of this material is at a higher mathematical level than the text. I realize that instructors
use this book with a wide array of audiences, and some more advanced design courses could possibly benefit from
including several of the supplemental text material topics. This material is in electronic form on the World Wide
Website for this book, located at www.wiley.com/college/montgomery.
Website
Current supporting material for instructors and students is available at the website www.wiley.com/college/
montgomery. This site will be used to communicate information about innovations and recommendations for
effectively using this text. The supplemental text material described above is available at the site, along with electronic
versions of data sets used for examples and homework problems, a course syllabus, and some representative student
term projects from the course at Arizona State University.
k
k
vi Preface
Acknowledgments
I express my appreciation to the many students, instructors, and colleagues who have used the eight earlier editions of
this book and who have made helpful suggestions for its revision. The contributions of Dr. Raymond H. Myers, Dr. G.
Geoffrey Vining, Dr. Brad Jones, Dr. Christine Anderson-Cook, Dr. Connie M. Borror, Dr. Scott Kowalski, Dr. Rachel
Silvestrini, Dr. Megan Olson Hunt, Dr. Dennis Lin, Dr. John Ramberg, Dr. Joseph Pignatiello, Dr. Lloyd S. Nelson, Dr.
Andre Khuri, Dr. Peter Nelson, Dr. John A. Cornell, Dr. Saeed Maghsoodloo, Dr. Don Holcomb, Dr. George C. Runger,
Dr. Bert Keats, Dr. Dwayne Rollier, Dr. Norma Hubele, Dr. Murat Kulahci, Dr. Cynthia Lowry, Dr. Russell G. Heikes,
Dr. Harrison M. Wadsworth, Dr. William W. Hines, Dr. Arvind Shah, Dr. Jane Ammons, Dr. Diane Schaub, Mr. Mark
Anderson, Mr. Pat Whitcomb, Dr. Pat Spagon, and Dr. William DuMouche were particularly valuable. My current
and former School Director and Department Chair, Dr. Ron Askin and Dr. Gary Hogg, have provided an intellectually
stimulating environment in which to work.
The contributions of the professional practitioners with whom I have worked have been invaluable. It is impossi-
ble to mention everyone, but some of the major contributors include Dr. Dan McCarville, Dr. Lisa Custer, Dr. Richard
Post, Mr. Tom Bingham, Mr. Dick Vaughn, Dr. Julian Anderson, Mr. Richard Alkire, and Mr. Chase Neilson of the
Boeing Company; Mr. Mike Goza, Mr. Don Walton, Ms. Karen Madison, Mr. Jeff Stevens, and Mr. Bob Kohm of
Alcoa; Dr. Jay Gardiner, Mr. John Butora, Mr. Dana Lesher, Mr. Lolly Marwah, Mr. Leon Mason of IBM; Dr. Paul
Tobias of IBM and Sematech; Ms. Elizabeth A. Peck of The Coca-Cola Company; Dr. Sadri Khalessi and Mr. Franz
Wagner of Signetics; Mr. Robert V. Baxley of Monsanto Chemicals; Mr. Harry Peterson-Nedry and Dr. Russell Boyles
of Precision Castparts Corporation; Mr. Bill New and Mr. Randy Schmid of Allied-Signal Aerospace; Mr. John M.
Fluke, Jr. of the John Fluke Manufacturing Company; Mr. Larry Newton and Mr. Kip Howlett of Georgia-Pacific; and
Dr. Ernesto Ramos of BBN Software Products Corporation.
k
k
Preface vii
I am indebted to Professor E. S. Pearson and the Biometrika Trustees, John Wiley & Sons, Prentice Hall, The
American Statistical Association, The Institute of Mathematical Statistics, and the editors of Biometrics for permission
to use copyrighted material. Dr. Lisa Custer and Dr. Dan McCorville did an excellent job of preparing the solutions
that appear in the Instructor’s Solutions Manual, and Dr. Cheryl Jennings provided effective and very helpful proof-
reading assistance. I am grateful to NASA, the Office of Naval Research, the Department of Defense, the National
Science Foundation, the member companies of the NSF/Industry/University Cooperative Research Center in Quality
and Reliability Engineering at Arizona State University, and the IBM Corporation for supporting much of my research
in engineering statistics and experimental design over many years.
DOUGLAS C. MONTGOMERY
TEMPE, ARIZONA
k k
k
k
Contents
Preface iii
1
Introduction 1
k k
1.1 Strategy of Experimentation 1
1.2 Some Typical Applications of Experimental Design 7
1.3 Basic Principles 11
1.4 Guidelines for Designing Experiments 13
1.5 A Brief History of Statistical Design 19
1.6 Summary: Using Statistical Techniques in Experimentation 20
1.7 Problems 21
2
Simple Comparative Experiments 23
2.1 Introduction 24
2.2 Basic Statistical Concepts 25
2.3 Sampling and Sampling Distributions 28
2.4 Inferences About the Differences in Means, Randomized Designs 33
2.4.1 Hypothesis Testing 33
2.4.2 Confidence Intervals 39
2.4.3 Choice of Sample Size 41
2.4.4 The Case Where 𝜎12 ≠ 𝜎22 44
2.4.5 The Case Where 𝜎12 and 𝜎22 Are Known 47
2.4.6 Comparing a Single Mean to a Specified Value 47
2.4.7 Summary 48
2.5 Inferences About the Differences in Means, Paired Comparison Designs 50
2.5.1 The Paired Comparison Problem 50
2.5.2 Advantages of the Paired Comparison Design 52
2.6 Inferences About the Variances of Normal Distributions 53
2.7 Problems 55
k
3
Experiments with a Single Factor: The Analysis of Variance 64
3.1 An Example 65
3.2 The Analysis of Variance 67
3.3 Analysis of the Fixed Effects Model 69
3.3.1 Decomposition of the Total Sum of Squares 69
3.3.2 Statistical Analysis 72
3.3.3 Estimation of the Model Parameters 76
3.3.4 Unbalanced Data 78
3.4 Model Adequacy Checking 78
3.4.1 The Normality Assumption 79
3.4.2 Plot of Residuals in Time Sequence 81
3.4.3 Plot of Residuals Versus Fitted Values 81
3.4.4 Plots of Residuals Versus Other Variables 86
3.5 Practical Interpretation of Results 86
3.5.1 A Regression Model 87
3.5.2 Comparisons Among Treatment Means 88
3.5.3 Graphical Comparisons of Means 88
3.5.4 Contrasts 89
3.5.5 Orthogonal Contrasts 92
3.5.6 Scheffé’s Method for Comparing All Contrasts 93
3.5.7 Comparing Pairs of Treatment Means 95
3.5.8 Comparing Treatment Means with a Control 98
k 3.6 Sample Computer Output 99 k
3.7 Determining Sample Size 103
3.7.1 Operating Characteristic and Power Curves 103
3.7.2 Confidence Interval Estimation Method 104
3.8 Other Examples of Single-Factor Experiments 105
3.8.1 Chocolate and Cardiovascular Health 105
3.8.2 A Real Economy Application of a Designed Experiment 107
3.8.3 Discovering Dispersion Effects 109
3.9 The Random Effects Model 111
3.9.1 A Single Random Factor 111
3.9.2 Analysis of Variance for the Random Model 112
3.9.3 Estimating the Model Parameters 113
3.10 The Regression Approach to the Analysis of Variance 119
3.10.1 Least Squares Estimation of the Model Parameters 120
3.10.2 The General Regression Significance Test 121
3.11 Nonparametric Methods in the Analysis of Variance 123
3.11.1 The Kruskal–Wallis Test 123
3.11.2 General Comments on the Rank Transformation 124
3.12 Problems 125
4
Randomized Blocks, Latin Squares, and Related Designs 135
4.1 The Randomized Complete Block Design 135
4.1.1 Statistical Analysis of the RCBD 137
4.1.2 Model Adequacy Checking 145
4.1.3 Some Other Aspects of the Randomized Complete Block Design 145
4.1.4 Estimating Model Parameters and the General Regression Significance Test 150
k
4.2 The Latin Square Design 153
4.3 The Graeco-Latin Square Design 160
4.4 Balanced Incomplete Block Designs 162
4.4.1 Statistical Analysis of the BIBD 163
4.4.2 Least Squares Estimation of the Parameters 167
4.4.3 Recovery of Interblock Information in the BIBD 169
4.5 Problems 171
5
Introduction to Factorial Designs 179
5.1 Basic Definitions and Principles 179
5.2 The Advantage of Factorials 182
5.3 The Two-Factor Factorial Design 183
5.3.1 An Example 183
5.3.2 Statistical Analysis of the Fixed Effects Model 186
5.3.3 Model Adequacy Checking 191
5.3.4 Estimating the Model Parameters 194
5.3.5 Choice of Sample Size 196
5.3.6 The Assumption of No Interaction in a Two-Factor Model 197
5.3.7 One Observation per Cell 198
5.4 The General Factorial Design 201
5.5 Fitting Response Curves and Surfaces 206
k 5.6 Blocking in a Factorial Design 215 k
5.7 Problems 220
6
The 2k Factorial Design 230
6.1 Introduction 230
6.2 The 22 Design 231
6.3 The 23 Design 240
6.4 The General 2k Design 252
6.5 A Single Replicate of the 2k Design 254
6.6 Additional Examples of Unreplicated 2k Designs 268
6.7 2k Designs are Optimal Designs 280
6.8 The Addition of Center Points to the 2k Design 285
6.9 Why We Work with Coded Design Variables 290
6.10 Problems 292
7
Blocking and Confounding in the 2k Factorial Design 308
7.1 Introduction 308
7.2 Blocking a Replicated 2k Factorial Design 309
7.3 Confounding in the 2k Factorial Design 311
7.4 Confounding the 2k Factorial Design in Two Blocks 311
7.5 Another Illustration of Why Blocking Is Important 319
7.6 Confounding the 2k Factorial Design in Four Blocks 320
k
7.7 Confounding the 2k Factorial Design in 2p Blocks 322
7.8 Partial Confounding 323
7.9 Problems 325
8
Two-Level Fractional Factorial Designs 328
8.1 Introduction 329
8.2 The One-Half Fraction of the 2k Design 329
8.2.1 Definitions and Basic Principles 329
8.2.2 Design Resolution 332
8.2.3 Construction and Analysis of the One-Half Fraction 332
8.3 The One-Quarter Fraction of the 2k Design 344
8.4 The General 2k−p Fractional Factorial Design 351
8.4.1 Choosing a Design 351
8.4.2 Analysis of 2k−p Fractional Factorials 354
8.4.3 Blocking Fractional Factorials 355
8.5 Alias Structures in Fractional Factorials and Other Designs 360
8.6 Resolution III Designs 362
8.6.1 Constructing Resolution III Designs 362
8.6.2 Fold Over of Resolution III Fractions to Separate Aliased Effects 364
8.6.3 Plackett–Burman Designs 367
8.7 Resolution IV and V Designs 376
k 8.7.1 Resolution IV Designs 376 k
8.7.2 Sequential Experimentation with Resolution IV Designs 377
8.7.3 Resolution V Designs 383
8.8 Supersaturated Designs 384
8.9 Summary 385
8.10 Problems 386
9
Additional Design and Analysis Topics for Factorial
and Fractional Factorial Designs 405
9.1 The 3k Factorial Design 406
9.1.1 Notation and Motivation for the 3k Design 406
9.1.2 The 32 Design 407
9.1.3 The 33 Design 408
9.1.4 The General 3k Design 413
9.2 Confounding in the 3k Factorial Design 413
9.2.1 The 3k Factorial Design in Three Blocks 413
9.2.2 The 3k Factorial Design in Nine Blocks 416
9.2.3 The 3k Factorial Design in 3p Blocks 417
9.3 Fractional Replication of the 3k Factorial Design 418
9.3.1 The One-Third Fraction of the 3k Factorial Design 418
9.3.2 Other 3k−p Fractional Factorial Designs 421
9.4 Factorials with Mixed Levels 422
9.4.1 Factors at Two and Three Levels 422
9.4.2 Factors at Two and Four Levels 424
9.5 Nonregular Fractional Factorial Designs 425
k
9.5.1 Nonregular Fractional Factorial Designs for 6, 7, and 8 Factors in 16 Runs 427
9.5.2 Nonregular Fractional Factorial Designs for 9 Through 14 Factors in 16 Runs 436
9.5.3 Analysis of Nonregular Fractional Factorial Designs 441
9.6 Constructing Factorial and Fractional Factorial Designs Using
an Optimal Design Tool 442
9.6.1 Design Optimality Criterion 443
9.6.2 Examples of Optimal Designs 443
9.6.3 Extensions of the Optimal Design Approach 453
9.7 Problems 454
10
Fitting Regression Models
(online at www.wiley.com/college/montgomery) 460
10.1 Introduction 461
10.2 Linear Regression Models 461
10.3 Estimation of the Parameters in Linear Regression Models 462
10.4 Hypothesis Testing in Multiple Regression 473
10.4.1 Test for Significance of Regression 473
10.4.2 Tests on Individual Regression Coefficients and Groups of Coefficients 475
10.5 Confidence Intervals in Multiple Regression 478
10.5.1 Confidence Intervals on the Individual Regression Coefficients 478
10.5.2 Confidence Interval on the Mean Response 478
k k
10.6 Prediction of New Response Observations 479
10.7 Regression Model Diagnostics 480
10.7.1 Scaled Residuals and PRESS 480
10.7.2 Influence Diagnostics 483
10.8 Testing for Lack of Fit 483
10.9 Problems 485
11
Response Surface Methods and Designs 489
11.1 Introduction to Response Surface Methodology 490
11.2 The Method of Steepest Ascent 492
11.3 Analysis of a Second-Order Response Surface 497
11.3.1 Location of the Stationary Point 497
11.3.2 Characterizing the Response Surface 499
11.3.3 Ridge Systems 505
11.3.4 Multiple Responses 506
11.4 Experimental Designs for Fitting Response Surfaces 511
11.4.1 Designs for Fitting the First-Order Model 511
11.4.2 Designs for Fitting the Second-Order Model 511
11.4.3 Blocking in Response Surface Designs 518
11.4.4 Optimal Designs for Response Surfaces 521
11.5 Experiments with Computer Models 535
11.6 Mixture Experiments 542
11.7 Evolutionary Operation 553
11.8 Problems 558
k
12
Robust Parameter Design and Process Robustness
Studies (online at www.wiley.com/college/montgomery) 569
12.1 Introduction 569
12.2 Crossed Array Designs 571
12.3 Analysis of the Crossed Array Design 573
12.4 Combined Array Designs and the Response Model Approach 576
12.5 Choice of Designs 582
12.6 Problems 585
13
Experiments with Random Factors 589
13.1 Random Effects Models 589
13.2 The Two-Factor Factorial with Random Factors 590
13.3 The Two-Factor Mixed Model 597
13.4 Rules for Expected Mean Squares 602
13.5 Approximate F-Tests 605
13.6 Some Additional Topics on Estimation of Variance Components 609
13.6.1 Approximate Confidence Intervals on Variance Components 609
13.6.2 The Modified Large-Sample Method 613
k k
13.7 Problems 615
14
Nested and Split-Plot Designs 618
14.1 The Two-Stage Nested Design 619
14.1.1 Statistical Analysis 619
14.1.2 Diagnostic Checking 624
14.1.3 Variance Components 626
14.1.4 Staggered Nested Designs 626
14.2 The General m-Stage Nested Design 628
14.3 Designs with Both Nested and Factorial Factors 630
14.4 The Split-Plot Design 634
14.5 Other Variations of the Split-Plot Design 640
14.5.1 Split-Plot Designs with More Than Two Factors 640
14.5.2 The Split-Split-Plot Design 645
14.5.3 The Strip-Split-Plot Design 649
14.6 Problems 650
15
Other Design and Analysis Topics
(online at www.wiley.com/college/montgomery) 656
15.1 Nonnormal Responses and Transformations 657
15.1.1 Selecting a Transformation: The Box–Cox Method 657
15.1.2 The Generalized Linear Model 659
k
15.2 Unbalanced Data in a Factorial Design 666
15.2.1 Proportional Data: An Easy Case 667
15.2.2 Approximate Methods 668
15.2.3 The Exact Method 670
15.3 The Analysis of Covariance 670
15.3.1 Description of the Procedure 671
15.3.2 Computer Solution 679
15.3.3 Development by the General Regression Significance Test 680
15.3.4 Factorial Experiments with Covariates 682
15.4 Repeated Measures 692
15.5 Problems 694
k
Other documents randomly have
different content
I can remember, that it is hard to see where it is going to
end. I must say that, though I would never force a young
person's inclinations, yet I do think the parents should have
something to say as to their children's settlement. However,
a person of discretion will find ways of managing such
matters and preventing uncomfortable entanglements.
No, he was not at all like Lord Percy; but, after all, I
liked his looks. Andrew had been about the world a good
deal for a man of his years, having been on two or three
long sea-voyages, and he was by no means as awkward as
young men of that age are apt to be.
"We must not keep the old man here," he said. "The
tide will be favorable for his escape by the day after to-
morrow, and an English ship will be waiting for him off the
shore. But first I would fain have one more celebration of
the Holy Supper with some of our poor friends. Heaven
knows when we shall have another chance. But what is this
I hear about the Sablots?"
I saw Andrew clinch his hand and set his teeth hard at
the idea of such care being needful; and indeed it was a
new care for him. Times were not very good in England just
then, but they were far better than with us.
The old man's face lighted up, and he took off his hat.
"I have a few for madame, and Marie Duclas has some,
I know."
"I fear she will not be happy," said Jeanne, wiping her
eyes. "I had lost so many before she came, and she was so
delicate in her childhood, that I was always more careful of
her than of David, who never gave me an hour's anxiety
since he was born, except on that unlucky day when he
went to see the procession."
"The poor Jeanne was too easy with her. She indulged
her far too much," said Marie Lebrun. "She took all the
hardest and most unpleasant work on herself, to spare
Lucille, and leave her time for her needlework and her fine
spinning. If she had had to work as hard as my girls, she
would not have had so much time to indulge her foolish
fancies."
The walk had been long and rough for us all, and
especially for my mother, and we were not sorry to see the
tower standing dark against the sky, and to meet the
challenge of our outposts; for at all our meetings we had
our sentinels and our pass-words.
CHAPTER VII.
A SUDDEN SUMMONS.
FOR about a fortnight or more after the departure of the
pastor we had a very quiet, pleasant time. The weather was
lovely, and we made long excursions out of doors. We
gathered apples and quinces, and hunted for herbs and
flowers, for Andrew was a good deal of an herbalist (a
botanist, I think they call it now, though I am sure herbalist
is the prettier word), and he was in correspondence with
some learned gentleman in London on the subject of plants.
He told me many things about flowers that I had never
known or dreamed of before, showing me the several parts
of the blossom, the leaves, and roots, by means of a pocket
magnifying-glass which he always carried about him.