Statistical methods in water resources Dennis R.

Helsel online version

Available at ebookfinal.com
( 4.4/5.0 ★ | 422 downloads )

https://ebookfinal.com/download/statistical-methods-in-water-
resources-dennis-r-helsel/
 Statistical methods in water resources Dennis R. Helsel Pdf
Download

EBOOK

Available Formats

■ PDF eBook Study Guide Ebook

EXCLUSIVE 2025 ACADEMIC EDITION – LIMITED RELEASE

Available Instantly Access Library

 We believe these products will be a great fit for you. Click
the link to download now, or visit ebookfinal
to discover even more!

Statistics for Censored Environmental Data Using Minitab

and R Second Edition Dennis R. Helsel(Auth.)

https://ebookfinal.com/download/statistics-for-censored-environmental-
data-using-minitab-and-r-second-edition-dennis-r-helselauth/

Water resources planning and management R Quentin Grafton

https://ebookfinal.com/download/water-resources-planning-and-
management-r-quentin-grafton/

Robust Statistical Methods with R 1st Edition Jana

Jureckova

https://ebookfinal.com/download/robust-statistical-methods-with-r-1st-
edition-jana-jureckova/

Managing Water Resources Methods and Tools for a Systems

Approach Slobodan P. Simonovic

https://ebookfinal.com/download/managing-water-resources-methods-and-
tools-for-a-systems-approach-slobodan-p-simonovic/
Water Resources Law 2nd edition. Edition Janice Gray

https://ebookfinal.com/download/water-resources-law-2nd-edition-
edition-janice-gray/

Water Resources Engineering 2nd Edition Larry W. Mays

https://ebookfinal.com/download/water-resources-engineering-2nd-
edition-larry-w-mays/

Water resources systems analysis 1st Edition Mohammad

Karamouz

https://ebookfinal.com/download/water-resources-systems-analysis-1st-
edition-mohammad-karamouz/

Water Resources Planning 4th Edition Andrew A. Dzurik

https://ebookfinal.com/download/water-resources-planning-4th-edition-
andrew-a-dzurik/

Experiment Design and Statistical Methods For Behavioural

and Social Research David R. Boniface

https://ebookfinal.com/download/experiment-design-and-statistical-
methods-for-behavioural-and-social-research-david-r-boniface/
Statistical methods in water resources Dennis R. Helsel
Digital Instant Download
Author(s): Dennis R. Helsel, Robert M. Hirsch
ISBN(s): 9780444814630, 0444814639
Edition: Pap/Dskt
File Details: PDF, 9.58 MB
Year: 2002
Language: english
Techniques of Water-Resources Investigations of the United States Geological Survey

Book 4, Hydrologic Analysis and Interpretation

Chapter A3

Statistical Methods
in Water Resources
By D.R. Helsel and R.M. Hirsch
U.S. DEPARTMENT OF THE INTERIOR
GALE A. NORTON, Secretary

U.S. GEOLOGICAL SURVEY

Charles G. Groat, Director

September 2002

The use of firm, trade, and brand names in this report is for identification purposes only and does
not constitute endorsement by the U.S. Geological Survey.

Publication available at:

http://water.usgs.gov/pubs/twri/twri4a3/
Table of Contents
Preface xi

Chapter 1 Summarizing Data 1

1.1 Characteristics of Water Resources Data 2
1.2 Measures of Location 3
1.2.1 Classical Measure -- the Mean 3
1.2.2 Resistant Measure -- the Median 5
1.2.3 Other Measures of Location 6
1.3 Measures of Spread 7
1.3.1 Classical Measures 7
1.3.2 Resistant Measures 8
1.4 Measures of Skewness 9
1.4.1 Classical Measure of Skewness 9
1.4.2 Resistant Measure of Skewness 10
1.5 Other Resistant Measures 10
1.6 Outliers 11
1.7 Transformations 12
1.7.1 The Ladder of Powers 12

Chapter 2 Graphical Data Analysis 17

2.1 Graphical Analysis of Single Data Sets 19
2.1.1 Histograms 19
2.1.2 Stem and Leaf Diagrams 20
2.1.3 Quantile Plots 22
2.1.4 Boxplots 24
2.1.5 Probability Plots 26
2.2 Graphical Comparisons of Two or More Data Sets 35
2.2.1 Histograms 35
2.2.2 Dot and Line Plots of Means, Standard Deviations 35
2.2.3 Boxplots 38
2.2.4 Probability Plots 40
2.2.5 Q-Q Plots 41
2.3 Scatterplots and Enhancements 45
ii

2.3.1 Evaluating Linearity 45

2.3.2 Evaluating Differences in Location on a Scatterplot 47
2.3.3 Evaluating Differences in Spread 50
2.4 Graphs for Multivariate Data 51
2.4.1 Profile Plots 51
2.4.2 Star Plots 53
2.4.3 Trilinear Diagrams 56
2.4.4 Plots of Principal Components 58
2.4.5 Other Multivariate Plots 59

Chapter 3 Describing Uncertainty 65

3.1 Definition of Interval Estimates 66
3.2 Interpretation of Interval Estimates 67
3.3 Confidence Intervals for the Median 70
3.3.1 Nonparametric Interval Estimate for the Median 70
3.3.2 Parametric Interval Estimate for the Median 73
3.4 Confidence Intervals for the Mean 74
3.4.1 Symmetric Confidence Interval for the Mean 75
3.4.2 Asymmetric Confidence Interval for the Mean 76
3.5. Nonparametric Prediction Intervals 76
3.5.1 Two-Sided Nonparametric Prediction Interval 77
3.5.2 One-Sided Nonparametric Prediction Interval 78
3.6 Parametric Prediction Intervals 80
3.6.1 Symmetric Prediction Interval 80
3.6.2 Asymmetric Prediction Intervals 80
3.7 Confidence Intervals for Percentiles (Tolerance Intervals) 82
3.7.1 Nonparametric Confidence Intervals for Percentiles 83
3.7.2 Nonparametric Tests for Percentiles 84
3.7.3 Parametric Confidence Intervals for Percentiles 88
3.7.4 Parametric Tests for Percentiles 90
3.8 Other Uses for Confidence Intervals 90
3.8.1 Implications of Non-Normality for Detection of Outliers 90
3.8.2 Implications of Non-Normality for Quality Control 91
3.8.3 Implications of Non-Normality for Sampling Design 93

Chapter 4 Hypothesis Tests 97

4.1 Classification of Hypothesis Tests 99
4.1.1 Classification Based on Measurement Scales 99
4.1.2 Classification Based on the Data Distribution 100
 iii

4.2 Structure of Hypothesis Tests 101

4.2.1 Choose the Appropriate Test 101
4.2.2 Establish the Null and Alternate Hypotheses 104
4.2.3 Decide on an Acceptable Error Rate α 106
4.2.4 Compute the Test Statistic from the Data 107
4.2.5 Compute the p-Value 108
4.2.6 Make the Decision to Reject H0 or Not 108
4.3 The Rank-Sum Test as an Example of Hypothesis Testing 109
4.4 Tests for Normality 113

Chapter 5 Differences Between Two Independent Groups 117

5.1 The Rank-Sum Test 118
5.1.1 Null and Alternate Hypotheses 118
5.1.2 Computation of the Exact Test 119
5.1.3 The Large Sample Approximation 121
5.1.4 The Rank Transform Approximation 123
5.2 The t-Test 124
5.2.1 Assumptions of the Test 124
5.2.2 Computation of the t-Test 125
5.2.3 Modification for Unequal Variances 125
5.2.4 Consequences of Violating the t-Test's Assumptions 127
5.3 Graphical Presentation of Results 128
5.3.1 Side-by-Side Boxplots 128
5.3.2 Q-Q Plots 129
5.4 Estimating the Magnitude of Differences Between Two Groups 131
5.4.1 The Hodges-Lehmann Estimator 131
5.4.2 Confidence Interval for ∆ ^ 132
5.4.3 Difference Between Mean Values 134
5.4.4 Confidence Interval for x − y 134

Chapter 6 Matched-Pair Tests 137

6.1 The Sign Test 138
6.1.1 Null and Alternate Hypotheses 138
6.1.2 Computation of the Exact Test 138
6.1.3 The Large Sample Approximation 141
6.2 The Signed-Rank Test 142
6.2.1 Null and Alternate Hypotheses 142
6.2.2 Computation of the Exact Test 143
6.2.3 The Large Sample Approximation 145
6.2.4 The Rank Transform Approximation 147
iv

6.3 The Paired t-Test 147

6.3.1 Assumptions of the Test 147
6.3.2 Computation of the Paired t-Test 148
6.4 Consequences of Violating Test Assumptions 149
6.4.1 Assumption of Normality (t-Test) 149
6.4.2 Assumption of Symmetry (Signed-Rank Test) 150
6.5 Graphical Presentation of Results 150
6.5.1 Boxplots 151
6.5.2 Scatterplots With X=Y Line 151
6.6 Estimating the Magnitude of Differences Between Two Groups 153
6.6.1 The Median Difference (Sign Test) 153
6.6.2 The Hodges-Lehmann Estimator (Signed-Rank Test) 153
6.6.3 Mean Difference (t-Test) 155

Chapter 7 Comparing Several Independent Groups 157

7.1 Tests for Differences Due to One Factor 159
7.1.1 The Kruskal-Wallis Test 159
7.1.2 Analysis of Variance (One Factor) 164
7.2 Tests for the Effects of More Than One Factor 169
7.2.1 Nonparametric Multi-Factor Tests 170
7.2.2 Multi-Factor Analysis of Variance -- Factorial ANOVA 170
7.3 Blocking -- The Extension of Matched-Pair Tests 181
7.3.1 Median Polish 182
7.3.2 The Friedman Test 187
7.3.3 Median Aligned-Ranks ANOVA 191
7.3.4 Parametric Two-Factor ANOVA Without Replication 193
7.4 Multiple Comparison Tests 195
7.4.1 Parametric Multiple Comparisons 196
7.4.2 Nonparametric Multiple Comparisons 200
7.5 Presentation of Results 202
7.5.1 Graphical Comparisons of Several Independent Groups 202
7.5.2 Presentation of Multiple Comparison Tests 205

Chapter 8 Correlation 209

8.1 Characteristics of Correlation Coefficients 210
8.1.1 Monotonic Versus Linear Correlation 210
8.2 Kendall's Tau 212
8.2.1 Computation 212
8.2.2 Large Sample Approximation 213
8.2.3 Correction for Ties 215
 v

8.3 Spearman's Rho 217

8.4 Pearson's r 218

Chapter 9 Simple Linear Regression 221

9.1 The Linear Regression Model 222
9.1.1 Assumptions of Linear Regression 224
9.2 Computations 226
9.2.1 Properties of Least Squares Solutions 227
9.3 Building a Good Regression Model 228
9.4 Hypothesis Testing in Regression 237
9.4.1 Test for Whether the Slope Differs from Zero 237
9.4.2 Test for Whether the Intercept Differs from Zero 238
9.4.3 Confidence Intervals on Parameters 239
9.4.4 Confidence Intervals for the Mean Response 240
9.4.5 Prediction Intervals for Individual Estimates of y 241
9.5 Regression Diagnostics 244
9.5.1 Measures of Outliers in the x Direction 246
9.5.2 Measures of Outliers in the y Direction 246
9.5.3 Measures of Influence 248
9.5.4 Measures of Serial Correlation 250
9.6 Transformations of the Response (y) Variable 252
9.6.1 To Transform or Not to Transform? 252
9.6.2 Consequences of Transformation of y 253
9.6.3 Computing Predictions of Mass (Load) 255
9.6.4 An Example 257
9.7 Summary Guide to a Good SLR Model 261

Chapter 10 Alternative Methods to Regression 265

10.1 Kendall-Theil Robust Line 266
10.1.1 Computation of the Line 266
10.1.2 Properties of the Estimator 267
10.1.3 Test of Significance 272
10.1.4 Confidence Interval for Theil Slope 273
10.2 Alternative Parametric Linear Equations 274
10.2.1 OLS of X on Y 275
10.2.2 Line of Organic Correlation 276
10.2.3 Least Normal Squares 278
10.2.4 Summary of the Applicability of OLS, LOC and LNS 280
10.3 Weighted Least Squares 280
10.4 Iteratively Weighted Least Squares 283
vi

10.5 Smoothing 285

10.5.1 Moving Median Smooths 285
10.5.2 LOWESS 287
10.5.3 Polar Smoothing 291

Chapter 11 Multiple Linear Regression 295

11.1 Why Use MLR? 296
11.2 MLR Model 296
11.3 Hypothesis Tests for Multiple Regression 297
11.3.1 Nested F Tests 297
11.3.2 Overall F Test 298
11.3.3 Partial F Tests 298
11.4 Confidence Intervals 299
11.4.1 Variance-Covariance Matrix 299
11.4.2 Confidence Intervals for Slope Coefficients 299
11.4.3 Confidence Intervals for the Mean Response 300
11.4.4 Prediction Intervals for an Individual y 300
11.5 Regression Diagnostics 300
11.5.1 Partial Residual Plots 301
11.5.2 Leverage and Influence 301
11.5.3 Multi-Collinearity 305
11.6 Choosing the Best MLR Model 309
11.6.1 Stepwise Procedures 310
11.6.2 Overall Measures of Quality 313
11.7 Summary of Model Selection Criteria 315
11.8 Analysis of Covariance 316
11.8.1 Use of One Binary Variable 316
11.8.2 Multiple Binary Variables 318

Chapter 12 Trend Analysis 323

12.1 General Structure of Trend Tests 324
12.1.1 Purpose of Trend Testing 324
12.1.2 Approaches to Trend Testing 325
12.2 Trend Tests With No Exogenous Variable 326
12.2.1 Nonparametric Mann-Kendall Test 326
12.2.2 Parametric Regression of Y on T 328
12.2.3 Comparison of Simple Tests for Trend 328
12.3 Accounting for Exogenous Variables 329
12.3.1 Nonparametric Approach 334
12.3.2 Mixed Approach 335
 vii

12.3.3 Parametric Approach 335

12.3.4 Comparison of Approaches 336
12.4 Dealing With Seasonality 337
12.4.1 The Seasonal Kendall Test 338
12.4.2 Mixture Methods 340
12.4.3 Multiple Regression With Periodic Functions 341
12.4.4 Comparison of Methods 342
12.4.5 Presenting Seasonal Effects 343
12.4.6 Differences Between Seasonal Patterns 344
12.5 Use of Transformations in Trend Studies 346
12.6 Monotonic Trend versus Two Sample (Step) Trend 348
12.7 Applicability of Trend Tests With Censored Data 352

Chapter 13 Methods for Data Below the Reporting Limit 357

13.1 Methods for Estimating Summary Statistics 358
13.1.1 Simple Substitution Methods 358
13.1.2 Distributional Methods 360
13.1.3 Robust Methods 362
13.1.4 Recommendations 362
13.1.5 Multiple Reporting Limits 364
13.2 Methods for Hypothesis Testing 366
13.2.1 Simple Substitution Methods 366
13.2.2 Distributional Test Procedures 367
13.2.3 Nonparametric Tests 367
13.2.4 Hypothesis Testing With Multiple Reporting Limits 369
13.2.5 Recommendations 370
13.3 Methods For Regression With Censored Data 371
13.3.1 Kendall's Robust Line Fit 371
13.3.2 Tobit Regression 371
13.3.3 Logistic Regression 372
13.3.4 Contingency Tables 373
13.3.5 Rank Correlation Coefficients 373
13.3.6 Recommendations 374

Chapter 14 Discrete Relationships 377

14.1 Recording Categorical Data 378
14.2 Contingency Tables (Both Variables Nominal) 378
14.2.1 Performing the Test for Independence 379
14.2.2 Conditions Necessary for the Test 381
14.2.3 Location of the Differences 382
14.3 Kruskal-Wallis Test for Ordered Categorical Responses 382
viii

14.3.1 Computing the Test 383

14.3.2 Multiple Comparisons 385
14.4 Kendall's Tau for Categorical Data (Both Variables Ordinal) 385
14.4.1 Kendall's τ b for Tied Data 385
14.4.2 Test of Significance for τ b 388
14.5 Other Methods for Analysis of Categorical Data 390

Chapter 15 Regression for Discrete Responses 393

15.1 Regression for Binary Response Variables 394
15.1.1 Use of Ordinary Least Squares 394
15.2 Logistic Regression 395
15.2.1 Important Formulae 395
15.2.2 Computation by Maximum Likelihood 396
15.2.3 Hypothesis Tests 397
15.2.4 Amount of Uncertainty Explained, R2 398
15.2.5 Comparing Non-Nested Models 398
15.3 Alternatives to Logistic Regression 402
15.3.1 Discriminant Function Analysis 402
15.3.2 Rank-Sum Test 402
15.4 Logistic Regression for More Than Two Response Categories 403
15.4.1 Ordered Response Categories 403
15.4.2 Nominal Response Categories 405

Chapter 16 Presentation Graphics 409

16.1 The Value of Presentation Graphics 410
16.2 Precision of Graphs 411
16.2.1 Color 412
16.2.2 Shading 413
16.2.3 Volume and Area 416
16.2.4 Angle and Slope 417
16.2.5 Length 420
16.2 6 Position Along Nonaligned Scales 421
16.2.7 Position Along an Aligned Scale 423
16.3 Misleading Graphics to be Avoided 423
16.3.1 Perspective 423
16.3.2 Graphs With Numbers 426
16.3.3 Hidden Scale Breaks 427
16.3.4 Overlapping Histograms 428

References 433
 ix

Appendix A Construction of Boxplots 451

Appendix B Tables 456

Appendix C Data Sets 468

Appendix D Answers to Exercises 469

Index 503
x
 xi

Preface P f
This book began as class notes for a course we teach on applied statistical methods to
hydrologists of the Water Resources Division, U. S. Geological Survey (USGS). It reflects our
attempts to teach statistical methods which are appropriate for analysis of water resources data.
As interest in this course has grown outside of the USGS, incentive grew to develop the material
into a textbook. The topics covered are those we feel are of greatest usefulness to the practicing
water resources scientist. Yet all topics can be directly applied to many other types of
environmental data.

This book is not a stand-alone text on statistics, or a text on statistical hydrology. For example,
in addition to this material we use a textbook on introductory statistics in the USGS training
course. As a consequence, discussions of topics such as probability theory required in a general
statistics textbook will not be found here. Derivations of most equations are not presented.
Important tables included in all general statistics texts, such as quantiles of the normal
distribution, are not found here. Neither are details of how statistical distributions should be
fitted to flood data -- these are adequately covered in numerous books on statistical hydrology.

We have instead chosen to emphasize topics not always found in introductory statistics
textbooks, and often not adequately covered in statistical textbooks for scientists and engineers.
Tables included here, for example, are those found more often in books on nonparametric
statistics than in books likely to have been used in college courses for engineers. This book
points the environmental and water resources scientist to robust and nonparametric statistics,
and to exploratory data analysis. We believe that the characteristics of environmental (and
perhaps most other 'real') data drive analysis methods towards use of robust and nonparametric
methods.

Exercises are included at the end of chapters. In our course, students compute each type of
analysis (t-test, regression, etc.) the first time by hand. We choose the smaller, simpler examples
for hand computation. In this way the mechanics of the process are fully understood, and
computer software is seen as less mysterious.

We wish to acknowledge and thank several other scientists at the U. S. Geological Survey for
contributing ideas to this book. In particular, we thank those who have served as the other
instructors at the USGS training course. Ed Gilroy has critiqued and improved much of the
material found in this book. Tim Cohn has contributed in several areas, particularly to the
sections on bias correction in regression, and methods for data below the reporting limit.
Richard Alexander has added to the trend analysis chapter, and Charles Crawford has
contributed ideas for regression and ANOVA. Their work has undoubtedly made its way into
this book without adequate recognition.
xii

Professor Ken Potter (University of Wisconsin) and Dr. Gary Tasker (USGS) reviewed the
manuscript, spending long hours with no reward except the knowledge that they have improved
the work of others. For that we are very grateful. We also thank Madeline Sabin, who carefully
typed original drafts of the class notes on which the book is based. As always, the responsibility
for all errors and slanted thinking are ours alone.

Dennis R. Helsel

Robert M. Hirsch

Reston, VA USA
June, 1991

Citations of trade names in this book are for reference purposes only, and do not reflect endorsement by the
authors or by the U. S. Geological Survey
Chapter 1
Summarizing Data

When determining how to appropriately analyze any collection of data, the first consideration
must be the characteristics of the data themselves. Little is gained by employing analysis
procedures which assume that the data possess characteristics which in fact they do not. The
result of such false assumptions may be that the interpretations provided by the analysis are
incorrect, or unnecessarily inconclusive. Therefore we begin this book with a discussion of the
common characteristics of water resources data. These characteristics will determine the
selection of appropriate data analysis procedures.

One of the most frequent tasks when analyzing data is to describe and summarize those data in
forms which convey their important characteristics. "What is the sulfate concentration one
might expect in rainfall at this location"? "How variable is hydraulic conductivity"? "What is
the 100 year flood" (the 99th percentile of annual flood maxima)? Estimation of these and
similar summary statistics are basic to understanding data. Characteristics often described
include: a measure of the center of the data, a measure of spread or variability, a measure of the
symmetry of the data distribution, and perhaps estimates of extremes such as some large or small
percentile. This chapter discusses methods for summarizing or describing data.

This first chapter also quickly demonstrates one of the major themes of the book -- the use of
robust and resistant techniques. The reasons why one might prefer to use a resistant measure,
such as the median, over a more classical measure such as the mean, are explained.
2 Statistical Methods in Water Resources

The data about which a statement or summary is to be made are called the population, or
sometimes the target population. These might be concentrations in all waters of an aquifer or
stream reach, or all streamflows over some time at a particular site. Rarely are all such data
available to the scientist. It may be physically impossible to collect all data of interest (all the
water in a stream over the study period), or it may just be financially impossible to collect them.
Instead, a subset of the data called the sample is selected and measured in such a way that
conclusions about the sample may be extended to the entire population. Statistics computed
from the sample are only inferences or estimates about characteristics of the population, such as
location, spread, and skewness. Measures of location are usually the sample mean and sample
median. Measures of spread include the sample standard deviation and sample interquartile
range. Use of the term "sample" before each statistic explicitly demonstrates that these only
estimate the population value, the population mean or median, etc. As sample estimates are far
more common than measures based on the entire population, the term "mean" should be
interpreted as the "sample mean", and similarly for other statistics used in this book. When
population values are discussed they will be explicitly stated as such.

1.1 Characteristics of Water Resources Data

Data analyzed by the water resources scientist often have the following characteristics:
1. A lower bound of zero. No negative values are possible.
2. Presence of 'outliers', observations considerably higher or lower than most of the data,
which infrequently but regularly occur. outliers on the high side are more common in water
resources.
3. Positive skewness, due to items 1 and 2. An example of a skewed distribution, the
lognormal distribution, is presented in figure 1.1. Values of an observation on the
horizontal axis are plotted against the frequency with which that value occurs. These
density functions are like histograms of large data sets whose bars become infinitely narrow.
Skewness can be expected when outlying values occur in only one direction.
4. Non-normal distribution of data, due to items 1 - 3 above. Figure 1.2 shows an important
symmetric distribution, the normal. While many statistical tests assume data follow a
normal distribution as in figure 1.2, water resources data often look more like figure 1.1. In
addition, symmetry does not guarantee normality. Symmetric data with more observations
at both extremes (heavy tails) than occurs for a normal distribution are also non-normal.
5. Data reported only as below or above some threshold (censored data). Examples include
concentrations below one or more detection limits, annual flood stages known only to be
lower than a level which would have caused a public record of the flood, and hydraulic
heads known only to be above the land surface (artesian wells on old maps).
6. Seasonal patterns. Values tend to be higher or lower in certain seasons of the year.
Summarizing Data 3

7. Autocorrelation. Consecutive observations tend to be strongly correlated with each other.

For the most common kind of autocorrelation in water resources (positive autocorrelation),
high values tend to follow high values and low values tend to follow low values.
8. Dependence on other uncontrolled variables. Values strongly covary with water discharge,
hydraulic conductivity, sediment grain size, or some other variable.

Methods for analysis of water resources data, whether the simple summarization methods such
as those in this chapter, or the more complex procedures of later chapters, should recognize
these common characteristics.

1.2 Measures of Location

The mean and median are the two most commonly-used measures of location, though they are
not the only measures available. What are the properties of these two measures, and when
should one be employed over the other?

1.2.1 Classical Measure -- the Mean

The mean ( X ) is computed as the sum of all data values X i , divided by the sample size n:
n Xi
X= ∑ n [1.1]
i=1
For data which are in one of k groups, equation [1.1] can be rewritten to show that the overall
mean depends on the mean for each group, weighted by the number of observations ni in each
group:
n ni
X = ∑ Xi n [1.2]
i=1

where X i is the mean for group i. The influence of any one observation Xj on the mean can be
seen by placing all but that one observation in one "group", or
(n − 1) 1
X = X ( j) + Xj•n .
n
1
= X( j )+ ( X( j )− X( j )) • n . [1.3]
where X ( j ) is the mean of all observations excluding Xj. Each observation's influence on the
overall mean X is (Xj − X ( j ) ), the distance between the observation and the mean excluding
that observation. Thus all observations do not have the same influence on the mean. An
'outlier' observation, either high or low, has a much greater influence on the overall mean X
than does a more 'typical' observation, one closer to its X ( j ) .
4 Statistical Methods in Water Resources

Figure 1.1 Density Function for a Lognormal Distribution

Figure 1.2 Density Function for a Normal Distribution

Summarizing Data 5

Another way of illustrating this influence is to realize that the mean is the balance point of the
data, when each point is stacked on a number line (figure 1.3a). Data points further from the
center exert a stronger downward force than those closer to the center. If one point near the
center were removed, the balance point would only need a small adjustment to keep the data set
in balance. But if one outlying value were removed, the balance point would shift dramatically
(figure 1.3b). This sensitivity to the magnitudes of a small number of points in the data set
defines why the mean is not a "resistant" measure of location. It is not resistant to changes in
the presence of, or to changes in the magnitudes of, a few outlying observations.

When this strong influence of a few observations is desirable, the mean is an appropriate
measure of center. This usually occurs when computing units of mass, such as the average
concentration of sediment from several samples in a cross-section. Suppose that sediment
concentrations closer to the river banks were much higher than those in the center. Waters
represented by a bottle of high concentration would exert more influence (due to greater mass
of sediment per volume) on the final concentration than waters of low or average concentration.
This is entirely appropriate, as the same would occur if the stream itself were somehow
mechanically mixed throughout its cross section.

Figure 1.3a The mean (triangle) as balance point of a data set.

Figure 1.3b Shift of the mean downward after removal of outlier.

1.2.2 Resistant Measure -- the Median

The median, or 50th percentile P0.50 , is the central value of the distribution when the data are
ranked in order of magnitude. For an odd number of observations, the median is the data point
which has an equal number of observations both above and below it. For an even number of
observations, it is the average of the two central observations. To compute the median, first
6 Statistical Methods in Water Resources

rank the observations from smallest to largest, so that x1 is the smallest observation, up to xn ,
the largest observation. Then

median ( P0.50 ) = X(n+1)/2 when n is odd, and

1
median ( P0.50 ) = 2 (X(n/2) + X(n/2)+1) when n is even. [1.4]

The median is only minimally affected by the magnitude of a single observation, being
determined solely by the relative order of observations. This resistance to the effect of a change
in value or presence of outlying observations is often a desirable property. To demonstrate the
resistance of the median, suppose the last value of the following data set (a) of 7 observations
were multiplied by 10 to obtain data set (b):

Example 1:
(a) 2 4 8 9 11 11 12 X = 8.1 P.50= 9
(b) 2 4 8 9 11 11 120 X = 23.6 P.50= 9
(7+1)
The mean increases from 8.1 to 23.6. The median, the 2 th or 4th lowest data point,
is unaffected by the change.

When a summary value is desired that is not strongly influenced by a few extreme observations,
the median is preferable to the mean. One such example is the chemical concentration one
might expect to find over many streams in a given region. Using the median, one stream with
unusually high concentration has no greater effect on the estimate than one with low
concentration. The mean concentration may be pulled towards the outlier, and be higher than
concentrations found in most of the streams. Not so for the median.

1.2.3 Other Measures of Location

Three other measures of location are less frequently used: the mode, the geometric mean, and
the trimmed mean. The mode is the most frequently observed value. It is the value having the
highest bar in a histogram. It is far more applicable for grouped data, data which are recorded
only as falling into a finite number of categories, than for continuous data. It is very easy to
obtain, but a poor measure of location for continuous data, as its value often depends on the
arbitrary grouping of those data.

The geometric mean (GM) is often reported for positively skewed data sets. It is the mean of
the logarithms, transformed back to their original units.
GM = exp ( Y ), where Yi = ln (Xi) [1.5]
x
(in this book the natural, base e logarithm will be abbreviated ln, and its inverse e abbreviated
exp(x) ). For positively skewed data the geometric mean is usually quite close to the median. In
fact, when the logarithms of the data are symmetric, the geometric mean is an unbiased estimate
Summarizing Data 7

of the median. This is because the median and mean logarithms are equal, as in figure 1.2. When
transformed back to original units, the geometric mean continues to be an estimate for the
median, but is not an estimate for the mean (figure 1.1).

Compromises between the median and mean are available by trimming off several of the lowest
and highest observations, and calculating the mean of what is left. Such estimates of location are
not influenced by the most extreme (and perhaps anomalous) ends of the sample, as is the mean.
Yet they allow the magnitudes of most of the values to affect the estimate, unlike the median.
These estimators are called "trimmed means", and any desirable percentage of the data may be
trimmed away. The most common trimming is to remove 25 percent of the data on each end --
the resulting mean of the central 50 percent of data is commonly called the "trimmed mean", but
is more precisely the 25 percent trimmed mean. A "0% trimmed mean" is the sample mean
itself, while trimming all but 1 or 2 central values produces the median. Percentages of trimming
should be explicitly stated when used. The trimmed mean is a resistant estimator of location, as
it is not strongly influenced by outliers, and works well for a wide variety of distributional shapes
(normal, lognormal, etc.). It may be considered a weighted mean, where data beyond the cutoff
'window' are given a weight of 0, and those within the window a weight of 1.0 (see figure 1.4).

Figure 1.4. Window diagram for the trimmed mean

1.3 Measures of Spread

It is just as important to know how variable the data are as it is to know their general center or
location. Variability is quantified by measures of spread.

1.3.1 Classical Measures

The sample variance, and its square root the sample standard deviation, are the classical
measures of spread. Like the mean, they are strongly influenced by outlying values.
n (X i −X ) 2
s =∑
2
sample variance [1.6]
i=1 (n −1)
8 Statistical Methods in Water Resources

s = s2 sample standard deviation [1.7]

They are computed using the squares of deviations of data from the mean, so that outliers
influence their magnitudes even more so than for the mean. When outliers are present these
measures are unstable and inflated. They may give the impression of much greater spread than
is indicated by the majority of the data set.

1.3.2 Resistant Measures

The interquartile range (IQR) is the most commonly-used resistant measure of spread. It
measures the range of the central 50 percent of the data, and is not influenced at all by the 25
percent on either end. It is therefore the width of the non-zero weight window for the trimmed
mean of figure 1.4.

The IQR is defined as the 75th percentile minus the 25th percentile. The 75th, 50th (median)
and 25th percentiles split the data into four equal-sized quarters. The 75th percentile (P.75), also
called the upper quartile, is a value which exceeds no more than 75 percent of the data and is
exceeded by no more than 25 percent of the data. The 25th percentile (P.25) or lower quartile is
a value which exceeds no more than 25 percent of the data and is exceeded by no more than 75
percent. Consider a data set ordered from smallest to largest: Xi, i =1,...n. Percentiles (Pj) are
computed using equation [1.8]

Pj = X(n+1)•j [1.8]

where n is the sample size of Xi, and

j is the fraction of data less than or equal to the percentile value (for the 25th, 50th
and 75th percentiles, j= .25, .50, and .75).

Non-integer values of (n+1)•j imply linear interpolation between adjacent values of X. For the
example 1 data set given earlier, n=7, and therefore the 25th percentile is X(7+1)•.25 or X2 = 4,
the second lowest observation. The 75th percentile is X6 , the 6th lowest observation, or 11.
The IQR is therefore 11−4 = 7.

One resistant estimator of spread other than the IQR is the Median Absolute Deviation, or
MAD. The MAD is computed by first listing the absolute value of all differences |d| between
each observation and the median. The median of these absolute values is then the MAD.
MAD (Xi) = median |di|, where di = Xi − median (Xi) [1.9]

Comparison of each estimate of spread for the Example 1 data set is as follows. When the last
value is changed from 12 to 120, the standard deviation increases from 3.8 to 42.7. The IQR
and the MAD remain exactly the same.
Another Random Scribd Document
with Unrelated Content
 PARCERE SUBJECTIS

"Stop! Stop! Harry," cried Alicia shrilly. "What are you doing? You'll
have to go to the house first."
"Shall I?" said Harry. "All right. Two thirty-five, be it noted."
The vehicle came to a standstill, and instantly clouds of vapour
rose from the horses.
"Virgil!" thought Edwin, gazing at the archway, which filled him
with sudden horror, like an obscenity misplaced.

II

Less than ten minutes later, he and Hilda and Alicia, together with
three strange men, stood under the archway. Events had followed
one another quickly, to Edwin's undoing. When the wagonette drew
up in the grounds of the Governor's house, Harry Hesketh had
politely indicated that for his horses he preferred the stables of a
certain inn down the road to any stables that hospitality might offer;
and he had driven off, Mrs. Rotherwas urging him to return without
any delay so that tennis might begin. The Governor had been called
from home, and in his absence a high official of the prison was
deputed to show the visitors through the establishment. This official
was the first of the three strange men; the other two were visitors.
Janet had said that she would not go over the prison, because she
meant to play tennis and wished not to tire herself. Alicia said kindly
that she at any rate would go with Hilda,--though she had seen it all
before, it was interesting enough to see again.
Edwin had thereupon said that he should remain with Janet. But
immediately Mrs. Rotherwas, whose reception of him had been full
of the most friendly charm, had shown surprise, if not pain. What,--
come to Princetown without inspecting the wonderful prison, when
the chance was there? Inconceivable! Edwin might in his blunt Five
Towns way have withstood Mrs. Rotherwas, but he could not
withstand Hilda, who, frowning, seemed almost ready to risk a public
altercation in order to secure his attendance. He had to yield. To
make a scene, even a very little one, in the garden full of light
dresses and polite suave voices would have been monstrous. He
thought of all that he had ever heard of the subjection of men to
women. He thought of Johnnie and of Mrs. Chris Hamson, who was
known for her steely caprices. And he thought also of Jimmie and of
the undesirable Mrs. Jimmie, who, it was said, had threatened to
love Jimmie no more unless he took her once a week without fail to
the theatre, whatever the piece, and played cards with her and two
of her friends on all the other nights of the week. He thought of men
as a sex conquered by the unscrupulous and the implacable, and in
this mood, superimposed on his mood of disgust at the mere sight of
the archway, he followed the high official and his train. Mrs.
Rotherwas's last words were that they were not to be long. But the
official said privately to the group that they must at any rate
approach the precincts of the prison with all ceremony, and he led
them proudly, with an air of ownership, round to the main entrance
where the wagonette had first stopped.
 A turnkey on the other side of the immense gates, using a
theatrical gesture, jangled a great bouquet of keys; the portal
opened, increasing the pride of the official, and the next moment
they were interned in the outer courtyard. The moor and all that it
meant lay unattainably beyond that portal. As the group slowly
crossed the enclosed space, with the grim façades of yellow-brown
buildings on each side and vistas of further gates and buildings in
front, the official and the two male visitors began to talk together
over the heads of Alicia and Hilda. The women held close to each
other, and the official kept upon them a chivalrous eye; the two
visitors were friends; Edwin was left out of the social scheme, and
lagged somewhat behind, like one who is not wanted but who
cannot be abandoned. He walked self-conscious, miserable,
resentful, and darkly angry. In one instant the three men had
estimated him, decided that he was not of their clan nor of any
related clan, and ignored him. Whereas the official and the two male
visitors, who had never met before, grew more and more friendly
each minute. One said that he did not know So-and-So of the Scots
Greys, but he knew his cousin Trevor of the Hussars, who had in fact
married a niece of his own. And then another question about
somebody else was asked, and immediately they were engaged in
following clues, as explorers will follow the intricate mouths of a
great delta and so unite in the main stream. They were happy.
Edwin did not seriously mind that; but what he did mind was
their accent--in those days termed throughout the Midlands "lah-di-
dah" (an onomatopoeic description), which, falsifying every vowel
sound in the language, and several consonants, magically created
around them an aura of utter superiority to the rest of the world. He
quite unreasonably hated them, and he also envied them, because
this accent was their native tongue, and because their clothes were
not cut like his, and because they were entirely at their ease.
Useless for the official to throw him an urbane word now and then;
neither his hate nor his constraint would consent to be alleviated;
the urbane words grew less frequent. Also Edwin despised them
because they were seemingly insensible to the tremendous horror of
the jail set there like an outrage in the midst of primitive and sane
Dartmoor. "Yes," their attitude said. "This is a prison, one of the
institutions necessary to the well being of society, like a workhouse
or an opera house,--an interesting sight!"
A second pair of iron gates were opened with the same
elaborate theatricality as the first, and while the operation was being
done the official, invigorated by the fawning of turnkeys, conversed
with Alicia, who during her short married life had acquired some
shallow acquaintance with the clans, and he even drew a reluctant
phrase from Hilda. Then, after another open space, came a third pair
of iron gates, final and terrific, and at length the party was under
cover, and even the sky of the moor was lost. Edwin, bored,
disgusted, shamed, and stricken, yielded himself proudly and
submissively to the horror of the experience.

III

Hilda had only one thought--would she catch sight of the innocent
prisoner? The party was now deeply engaged in a system of
corridors and stairways. The official had said that as the tour of
inspection was to be short he would display to them chiefly the
modern part of the prison. So far not a prisoner had been seen, and
scarcely a warder. The two male visitors were scientifically interested
in the question of escapes. Did prisoners ever escape?
"Never!" said the official, with satisfaction.
"Impossible, I suppose. Even when they're working out on the
moor? Warders are pretty good shots, eh?"
"Practically impossible," said the official. "But there is one way."
He looked up the stairway on whose landing they stood, and down
the stairway, and cautiously lowered his voice. "Of course what I tell
you is confidential. If one of our Dartmoor fogs came on suddenly,
and kind friends outside had hidden a stock of clothes and food in
an arranged spot, then theoretically--I say, theoretically--a man
might get away. But nobody ever has done."
"I suppose you still have the silent system?"
The official nodded.
"Absolutely?"
"Absolutely."
"How awful it must be!" said Alicia, with a nervous laugh.
The official shrugged his shoulders, and the other two males
murmured reassuring axioms about discipline.
They emerged from the stairway into a colossal and resounding
iron hall. Round the emptiness of this interior ran galleries of
perforated iron protected from the abyss by iron balustrades. The
group stood on the second of the galleries from the stony floor, and
there were two galleries above them. Far away, opposite, a glint of
sunshine had feloniously slipped in, transpiercing the gloom, and it
lighted a series of doors. There was a row of these doors along
every gallery. Each had a peep-hole, a key-hole and a number. The
longer Hilda regarded, the more nightmarishly numerous seemed
the doors. The place was like a huge rabbit-hutch designed for the
claustration of countless rabbits. Across the whole width and length
of the hall, and at the level of the lowest gallery, was stretched a
great net.
"To provide against suicides?" suggested one of the men.
"Yes," said the official.
"A good idea."
When the reverberation of the words had ceased, a little silence
ensued. The ear listened vainly for the slightest sound. In the silence
the implacability of granite walls and iron reticulations reigned over
the accursed vision, stultifying the soul.
"Are these cells occupied?" asked Alicia timidly.
"Not yet, Mrs. Hesketh. It's too soon. A few are."
Hilda thought:
"He may be here,--behind one of those doors." Her heart was
liquid with compassion and revolt. "No," she assured herself. "They
must have taken him away already. It's impossible he should be
here. He's innocent."
"Perhaps you would like to see one of the cells?" the official
suggested.
A warder appeared, and, with the inescapable jangle of keys,
opened a door. The party entered the cell, ladies first, then the
official and his new acquaintances; then Edwin, trailing. The cell was
long and narrow, fairly lofty, bluish-white colour, very dimly lighted
by a tiny grimed window high up in a wall of extreme thickness. The
bed lay next the long wall; except the bed, a stool, a shelf, and some
utensils, there was nothing to furnish the horrible nakedness of the
cell. One of the visitors picked up an old book from the shelf. It was
a Greek Testament. The party seemed astonished at this evidence of
culture among prisoners, of the height from which a criminal may
have fallen.
The official smiled.
"They often ask for such things on purpose," said he. "They
think it's effective. They're very naïve, you know, at bottom."
"This very cell may be his cell," thought Hilda. "He may have
been here all these months, years, knowing he was innocent. He
may have thought about me in this cell." She glanced cautiously at
Edwin, but Edwin would not catch her eye.
They left. On the way to the workshops, they had a glimpse of
the old parts of the prison, used during the Napoleonic wars,
incredibly dark, frowsy, like catacombs.
"We don't use this part--unless we're very full up," said the
official, and he contrasted it with the bright, spacious, healthy
excellences of the hall which they had just quitted, to prove that
civilisation never stood still.
And then suddenly, at the end of a passage, a door opened and
they were in the tailors' shop, a large irregular apartment full of a
strong stench and of squatted and grotesque human beings. The
human beings, for the most part, were clothed in a peculiar brown
stuff, covered with broad arrows. The dress consisted of a short
jacket, baggy knickerbockers, black stockings, and coloured shoes.
Their hair was cut so short that they had the appearance of being
bald, and their great ears protruded at a startling angle from the
sides of those smooth heads. They were of every age, yet they all
looked alike, ridiculous, pantomimic, appalling. Some gazed with
indifference at the visitors; others seemed oblivious of the entry.
They all stitched on their haunches, in the stench, under the
surveillance of eight armed warders in blue.
"How many?" asked the official mechanically.
"Forty-nine, sir," said a warder.
And Hilda searched their loathsome and vapid faces for the face
of George Cannon. He was not there. She trembled,--whether with
relief or with disappointment she knew not. She was agonised, but
in her torture she exulted that she had come.
No comment had been made in the workshop, the official
having hinted that silence was usual on such occasions. But in a kind
of antechamber--one of those amorphous spaces, serving no
purpose and resembling nothing, which are sometimes to be found
between definable rooms and corridors in a vast building imperfectly
planned--the party halted in the midst of a discussion as to
discipline. The male visitors, except Edwin, showed marked
intelligence and detachment; they seemed to understand
immediately how it was that forty-nine ruffians could be trusted to
squat on their thighs and stitch industriously and use scissors and
other weapons for hours without being chained to the ground; they
certainly knew something of the handling of men. The official,
triumphant, stated that every prisoner had the right of personal
appeal to the Governor every day.
"They come with their stories of grievances," said he, tolerant
and derisive.
"Which often aren't true?"
 "Which are never true," said the official quietly. "Never! They
are always lies--always! ... Shows the material we have to deal
with!" He gave a short laugh.
"Really!" said one of the men, rather pleased and excited by this
report of universal lying.
"I suppose," Edwin blurted out, "you can tell for certain when
they aren't speaking the truth?"
Everybody looked at him surprised, as though the dumb had
spoken. The official's glance showed some suspicion of sarcasm and
a tendency to resent it.
"We can," he answered shortly, commanding his features to a
faint smile. "And now I wonder what Mrs. Rotherwas will be saying if
I don't restore you to her." It was agreed that regard must be had
for Mrs. Rotherwas's hospitable arrangements, though the prison
was really very interesting and would repay study.
They entered a wide corridor--one of two that met at right-
angles in the amorphous space--leading in the direction of the chief
entrance. From the end of this corridor a file of convicts was
approaching in charge of two warders with guns. The official offered
no remark, but held on. Hilda, falling back near to Edwin in the
procession, was divided between a dreadful fear and a hope equally
dreadful. Except in the tailors' shop, these were the only prisoners
they had seen, and they appeared out of place in the half-freedom
of the corridor; for nobody could conceive a prisoner save in a cell or
shop, and these were moving in a public corridor, unshackled.
Then she distinguished George Cannon among them. He was
the third from the last. She knew him by his nose and the shape of
his chin, and by his walk, though there was little left of his proud
walk in the desolating, hopeless prison-shuffle which was the gait of
all six convicts. His hair was iron-grey. All these details she could see
and be sure of in the distance of the dim corridor. She no longer had
a stomach; it had gone, and yet she felt a horrible nausea.
She cried out to herself:
"Why did I come? Why did I come? I am always doing these
mad things. Edwin was right. Why do I not listen to him?"
The party of visitors led by the high official, and the file of
convicts in charge of armed warders, were gradually approaching
one another in the wide corridor. It seemed to Hilda that a fearful
collision was imminent, and that something ought to be done. But
nobody among the visitors did anything or seemed to be disturbed.
Only they had all fallen silent; and in the echoing corridor could be
heard the firm steps of the male visitors accompanying the delicate
tripping of the women, and the military tramp of the warders with
the confused shuffling of the convicts.
"Has he recognised me?" thought Hilda, wildly.
She hoped that he had and that he had not. She recalled with
the most poignant sorrow the few days of their union, their hours of
intimacy, his kisses, her secret realisation of her power over him,
and of his passion. She wanted to scream:
"That man there is as innocent as any of you, and soon the
whole world will know it! He never committed any crime except that
of loving me too much. He could not do without me, and so I was
his ruin. It is horrible that he should be here in this hell. He must be
set free at once. The Home Secretary knows he is innocent, but they
are so slow. How can anyone bear that he should stop here one
instant longer?"
 But she made no sound. The tremendous force of an ancient
and organised society kept her lips closed and her feet in a line with
the others. She thought in despair:
"We are getting nearer, and I cannot meet him. I shall drop."
She glanced at Edwin, as if for help, but Edwin was looking straight
ahead.
Then a warder, stopping, ejaculated with the harsh brevity of a
drill-serjeant:
"Halt!"
The file halted.
"Right turn!"
The six captives turned, with their faces close against the wall
of the corridor, obedient, humiliated, spiritless, limp, stooping. Their
backs presented the most ridiculous aspect; all the calculated
grotesquerie of the surpassingly ugly prison uniform was
accentuated as they stood thus, a row of living scarecrows, who
knew that they had not the right even to look upon free men. Every
one of them except George Cannon had large protuberant ears that
completed the monstrosity of their appearance.
The official gave his new acquaintances a satisfied glance, as if
saying:
"That is the rule by which we manage these chance
encounters."
The visitors went by in silence, instinctively edging away from
the captives. And as she passed, Hilda lurched very heavily against
Edwin, and recovered herself. Edwin seized her arm near the
shoulder, and saw that she was pale. The others were in front.
Behind them they could hear the warder:
 "Left turn! March!"
And the shuffling and the tramping recommenced.

IV

In the garden of the Governor's house tennis had already begun

when the official brought back his convoy. Young Truscott and Mrs.
Rotherwas were pitted against Harry Hesketh and a girl of eighteen
who possessed a good wrist but could not keep her head. Harry was
watching over his partner, quietly advising her upon the ruses of the
enemy, taking the more difficult strokes for her, and generally
imparting to her the quality which she lacked. Harry was fully
engaged; the whole of his brain and body was at strain; he let
nothing go by; he missed no chance, and within the laws of the
game he hesitated at no stratagem. And he was beating young
Truscott and Mrs. Rotherwas, while an increasing and polite
audience looked on. To the entering party, the withdrawn scene, lit
by sunshine, appeared as perfect as a stage-show, with its trees,
lawn, flowers, toilettes, the flying balls, the grace of the players, and
the grey solidity of the governor's house in the background.
Alicia ran gawkily to Janet, who had got a box of chocolates
from somewhere, and one of the men followed her, laughing. Hilda
sat apart; she was less pale. Edwin remained cautiously near her. He
had not left her side since she lurched against him in the corridor. He
knew; he had divined that that which he most feared had come to
pass,--the supreme punishment of Hilda's morbidity. He had not
definitely recognised George Cannon, for he was not acquainted with
him, and in the past had only once or twice by chance caught sight
of him in the streets of Bursley or Turnhill. But he had seen among
the six captives one who might be he, and who certainly had
something of the Five Towns look. Hilda's lurch told him that by
vindictiveness of fate George Cannon was close to them.
He had ignored his own emotion. The sudden transient weight
of Hilda's body had had a strange moral effect upon him. "This," he
thought, "is the burden I have to bear. This, and not lithography, nor
riches, is my chief concern. She depends on me. I am all she has to
stand by." The burden with its immense and complex responsibilities
was sweet to his inmost being; and it braced him and destroyed his
resentment against her morbidity. His pity was pure. He felt that he
must live more nobly--yes, more heroically--than he had been living;
that all irritable pettiness must drop away from him, and that his
existence in her regard must have simplicity and grandeur. The
sensation of her actual weight stayed with him. He had not spoken
to her; he dared not; he had scarcely met her eyes; but he was
ready for any emergency. Every now and then, in the garden, Hilda
glanced over her shoulder at the house, as though her gaze could
pierce the house and see the sinister prison beyond.
The set ended, to Harry Hesketh's satisfaction; and, another set
being arranged, he and Mrs. Rotherwas, athletic in a short skirt and
simple blouse, came walking, rather flushed and breathless, round
the garden with one or two others, including Harry's late partner.
The conversation turned upon the great South Wales colliery strike
against a proposed reduction of wages. Mrs. Rotherwas' husband
was a colliery proprietor near Monmouth, and she had just received
a letter from him. Everyone sympathised with her and her husband,
and nobody could comprehend the wrongheadedness of the miners,
except upon the supposition that they had been led away by
mischievous demagogues. As the group approached, the timid young
girl, having regained her nerve, was exclaiming with honest
indignation: "The leaders ought to be shot, and the men who won't
go down the pits ought to be forced to go down and made to work."
And she picked at fluff on her yellow frock. Edwin feared an uprising
from Hilda, but naught happened. Mrs. Rotherwas spoke about tea,
though it was rather early, and they all, Hilda as well, wandered to a
large yew tree under which was a table; through the pendant
branches of the tree the tennis could be watched as through a
screen.
The prison clock tolled the hour over the roofs of the house, and
Mrs. Rotherwas gave the definite signal for refreshments.
"You're exhausted," she said teasingly to Harry.
"You'll see," said Harry.
"No," Mrs. Rotherwas delightfully relented. "You're a dear, and I
love to watch you play. I'm sure you could give Mr. Truscott half
fifteen."
"Think so?" said Harry, pleased, and very conscious that he was
living fully.
"You see what it is to have an object in life, Hesketh," Edwin
remarked suddenly.
Harry glanced at him doubtfully, and yet with a certain
ingenuous admiration. At the same time a white ball rolled near the
tree. He ducked under the trailing branches, returned the ball, and
moved slowly towards the court.
 "Alicia tells me you're very old friends of theirs," said Mrs.
Rotherwas, agreeably, to Hilda.
Hilda smiled quietly.
"Yes, we are, both of us."
Who could have guessed, now, that her condition was not
absolutely normal?
"Charming people, aren't they, the Heskeths?" said Mrs.
Rotherwas. "Perfectly charming. They're an ideal couple. And I do
like their house, it's so deliciously quaint, isn't it, Mary?"
"Lovely," agreed the young girl.
It was an ideal world, full of ideal beings.
Soon after tea the irresistible magnetism of Alicia's babies drew
Alicia off the moor, and with her the champion player, Janet, Hilda
and Edwin. Mrs. Rotherwas let them go with regret, adorably
expressed. Harry would have liked to stay, but on the other hand he
was delightfully ready to yield to Alicia.

On arriving at Tavy Mansion Hilda announced that she should lie

down. She told Edwin, in an exhausted but friendly voice, that she
needed only rest, and he comprehended, rightly, that he was to
leave her. Not a word was said between them as to the events within
the prison. He left her, and spent the time before dinner with Harry
Hesketh, who had the idea of occupying their leisure with a short
game of bowls, for which it was necessary to remove the croquet
hoops.
 Hilda undressed and got into bed. Soon afterwards both Alicia,
with an infant, and Janet came to see her. Had Janet been alone,
Hilda might conceivably in her weakness have surrendered the
secret to her in exchange for that soft and persuasive sympathy of
which Janet was the mistress, but the presence of Alicia made a
confidence impossible, and Hilda was glad. She plausibly fibbed to
both sisters, and immediately afterwards the household knew that
Hilda would not appear at dinner. There was not the slightest alarm
or apprehension, for the affair explained itself in the simplest way,--
Hilda had had a headache in the morning, and had been wrong to
go out; she was now merely paying for the indiscretion. She would
be quite recovered the next day. Alicia whispered a word to her
husband, who, besides, was not apt easily to get nervous about
anything except his form at games. Edwin also, with his Five Towns
habit of mind, soberly belittled the indisposition. The household
remained natural and gay. When Edwin went upstairs to prepare for
dinner, moving very quietly, his wife had her face towards the wall
and away from the light. He came round the bed to look at her.
"I'm all right," she murmured.
"Want nothing at all?" he asked, with nervous gruffness.
She shook her head.
Very impatiently she awaited his departure, exasperated more
than she had ever been by his precise deliberation over certain
details of his toilet. As soon as he was gone she began to cry; but
the tears came so gently from her eyes that the weeping was as
passive, as independent of volition, as the escape of blood from a
wound.
 She had a grievance against Edwin. At the crisis in the prison
she had blamed herself for not submitting to his guidance, but now
she had reacted against all such accusations, and her grievance
amounted to just an indictment of his commonsense, his quietude,
his talent for keeping out of harm's way, his lack of violent impulses,
his formidable respectability. She was a rebel; he was not. He would
never do anything wrong, or even perilous. Never, never would he
find himself in need of a friend's help. He would always direct his
course so that society would protect him. He was a firm part of the
structure of society; he was the enemy of impulses. When he
foresaw a danger, the danger was always realised: she had noticed
that, and she resented it. He was infinitely above the George
Cannons of the world. He would be incapable of bigamy, incapable
of being caught in circumstances which could bring upon him
suspicion of any crime whatever. Yet for her the George Cannons
had a quality which he lacked, which he could never possess, and
which would have impossibly perfected him--a quality heroic, foolish,
martyr-like! She was almost ready to decide that his complete social
security was due to cowardice and resulted in self-righteousness! ...
Could he really feel pity as she felt it, for the despised and rejected,
and a hatred of injustice equal to hers?
These two emotions were burning her up. Again and again,
ceaselessly, her mind ran round the circle of George Cannon's
torture and the callousness of society. He had sinned, and she had
loathed him; but both his sin and her loathing were the fruit of
passion. He had been a proud man, and she had shared his pride;
now he was broken, unutterably humiliated, and she partook of his
humiliation. The grotesque and beaten animal in the corridor was all
that society had left of him who had once inspired her to acts of
devotion, who could make her blush, and to satisfy whom she would
recklessly spend herself. The situation was intolerable, and yet it had
to be borne. But surely it must be ended! Surely at the latest on the
morrow the prisoner must be released, and soothed and reinstated!
... Pardoned? No! A pardon was an insult, worse than an insult. She
would not listen to the word. Society might use it for its own
purposes; but she would never use it. Pardon a man after
deliberately and fiendishly achieving his ruin? She could have
laughed.
Exhaustion followed, tempering emotion and reducing it to a
profound despairing melancholy that was stirred at intervals by
frantic revolt. The light failed. The windows became vague silver
squares. Outside fowls clucked, a horse's hoof clattered on stones;
servants spoke to each other in their rough, good-natured voices.
The peace of the world had its effect on her, unwilling though she
was. Then there was a faint tap at the door. She made no reply, and
shut her eyes. The door gently opened, and someone tripped
delicately in. She heard movements at the washstand.... One of the
maids. A match was struck. The blinds were stealthily lowered, the
curtains drawn; garments were gathered together, and at last the
door closed again.
She opened her eyes. The room was very dimly illuminated. A
night-light, under a glass hemisphere of pale rose, stood on the
dressing-table. By magic, order had been restored; a glinting copper
ewer of hot water stood in the whiteness of the basin with a towel
over it; the blue blinds, revealed by the narrowness of the red
curtains, stirred in the depths of the windows; each detail of the
chamber was gradually disclosed, and the chamber was steeped in
the first tranquillity of the night. Not a sound could be heard.
Through the depths of her bitterness, there rose slowly the
sensation of the beauty of existence even in its sadness....
A long time afterwards it occurred to her in the obscurity that
the bed was tumbled. She must have turned over and over. The bed
must be arranged before Edwin came. He had to share it. After all,
he had committed no fault; he was entirely innocent. She and fate
between them had inflicted these difficulties and these solicitudes
upon him. He had said little or nothing, but he was sympathetic.
When she had stumbled against him she had felt his upholding
masculine strength. He was dependable, and would be dependable
to the last. The bed must be creaseless when he came; this was the
least she could do. She arose. Very faintly she could descry her
image in the mirror of the great wardrobe--a dishevelled image.
Forgetting the bed, she bathed her face, and, unusually, took care to
leave the washstand as tidy as the maid had left it. Then, having
arranged her hair, she set about the bed. It was not easy for one
person unaided to make a wide bed. Before she had finished she
heard footsteps outside the door. She stood still. Then she heard
Edwin's voice:
"Don't trouble, thanks. I'll take it in myself."
He entered, carrying a tray, and shut the door, and instantly she
busied herself once more with the bed.
"My poor girl," he said with quiet kindliness, "what are you
doing?"
"I'm just putting the bed to rights," she answered, and almost
with a single movement she slid back into the bed. "What have you
got there?"
"I thought I'd ask for some tea for you," he said. "Nearly the
whole blessed household wanted to come and see you, but I
wouldn't have it."
She could not say: "It's very nice of you." But she said, simply
to please him: "I should like some tea."
He put the tray on the dressing-table; then lit three candles,
two on the dressing-table and one on the night-table, and brought
the tray to the night-table.
He himself poured out the tea, and offered the cup. She raised
herself on an elbow.
"Did you recognise him?" she muttered suddenly, after she had
blown on the tea to cool it.
Under ordinary conditions Edwin would have replied to such an
unprepared question with another, petulant and impatient:
"Recognise who?" pretending that he did not understand the
allusion. But now he made no pretences.
"Not quite," he said. "But I knew at once. I could see which of
them it must be."
The subject at last opened between them, Hilda felt an
extraordinary solace and relief. He stood by the bedside, in black,
with a great breastplate of white, his hair rough, his hands in his
pockets. She thought he had a fine face; she thought of him as, at
such a time, her superior; she wanted powerfully to adopt his
attitude, to believe in everything he said. They were talking together
in safety, quietly, gravely, amicably, withdrawn and safe in the
strange house--he benevolent and assuaging and comprehending,
she desiring the balm which he could give. It seemed to her that
they had never talked to each other in such tones.
"Isn't it awful--awful?" she exclaimed.
"It is," said Edwin, and added carefully, tenderly: "I suppose he
is innocent."
She might have flown at him: "That's just like you--to assume
he isn't!" But she replied:
"I'm quite sure of it. I say--I want you to read all the letters I've
had from Mrs. Cannon. I've got them here. They're in my bag there.
Read them now. Of course I always meant to show them to you."
"All right," he agreed, drew a chair to the dressing-table where
the bag was, found the letters, and read them. She waited, as he
read one letter, put it down, read another, laid it precisely upon the
first one, with his terrible exactitude and orderliness, and so on
through the whole packet.
"Yes," said he at the end, "I should say he's innocent this time,
right enough."
"But something ought to be done!" she cried. "Don't you think
something ought to be done, Edwin?"
"Something has been done. Something is being done."
"But something else!"
He got up and walked about the room.
"There's only one thing to be done," he said.
He came towards her, and stood over her again, and the candle
on the night-table lighted his chin and the space between his
eyelashes and his eyebrows. He timidly touched her hair, caressing
it. They were absolutely at their ease together in the intimacy of the
bedroom. In her brief relations with George Cannon there had not
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookfinal.com