Deocument - 631full Download Book Multivariate Data Analysis PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 41

Multivariate Data Analysis - eBook PDF

Visit to download the full and correct content document:


https://ebooksecure.com/download/multivariate-data-analysis-ebook-pdf/
EIGHTH EDITION

MULTIVARIATE
DATA
ANALYSIS
Joseph F. Hair Jr., William C. Black,
Barry J. Babin, Rolph E. Anderson
Multivariate Data
Analysis
EIGHTH EDITION

Joseph F. Hair, Jr.


University of South Alabama

William C. Black
Louisiana State University

Barry J. Babin
Louisiana Tech University

Rolph E. Anderson
Drexel University

Australia • Brazil • Mexico • South Africa • Singapore • United Kingdom • United States

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
This is an electronic version of the print textbook. Due to electronic rights restrictions,
some third party content may be suppressed. Editorial review has deemed that any suppressed
content does not materially affect the overall learning experience. The publisher reserves the right
to remove content from this title at any time if subsequent rights restrictions require it. For
valuable information on pricing, previous editions, changes to current editions, and alternate
formats, please visit www.cengage.com/highered to search by ISBN#, author, title, or keyword for
materials in your areas of interest.

Important Notice: Media content referenced within the product description or the product
text may not be available in the eBook version.

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Multivariate Data Analysis, © 2019, Cengage Learning EMEA
Eighth Edition
WCN: 02-300
ALL RIGHTS RESERVED. No part of this work covered by the
Joseph F. Hair Jr, William C. Black,

copyright herein may be reproduced or distributed in any form


Barry J. Babin, Rolph E. Anderson

or by any means, except as permitted by U.S. copyright law,


Publisher: Annabel Ainscow
without the prior written permission of the copyright owner.
List Manager: Jenny Grene

Marketing Manager: Sophie Clarke

Content Project Manager: Melissa Beavis For product information and technology assistance, contact us at

Manufacturing Manager: Eyvett Davis


emea.info@cengage.com

Typesetter: SPi Global For permission to use material from this text or product and for
Text Design: SPi Global permission queries, email emea.permissions@cengage.com

Cover Designer: Simon Levy Associates

Cover Images: Istockphoto/liuzishan


British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British


Library. This book was previously published by Pearson
Education, Inc.

ISBN: 978-1-4737-5654-0

Cengage Learning, EMEA


Cheriton House, North Way
Andover, Hampshire, SP10 5BE
United Kingdom

Cengage Learning is a leading provider of customized learning


solutions with employees residing in nearly 40 diferent
countries and sales in more than 125 countries around the
world. Find your local representative at: www.cengage.co.uk.

Cengage Learning products are represented in Canada by


Nelson Education, Ltd.

For your course and learning solutions, visit www.cengage.co.uk

Purchase any of our products at your local college store or at


our preferred online store www.cengagebrain.com.

Printed in China by RR Donnelley


Print Number: 01 Print Year: 2018

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
To my family, and particularly my wife Dale
—Joseph F. Hair, Jr., Mobile, Alabama

To Deb, Steve, Emily and especially Alden, my granddaughter,


for their love and support in this new stage of my career
—William C. Black, Austin, TX

For Laurie, Amie, and James, and my mother Barbara


—Barry J. Babin, Choudrant, LA

To Rachel and Stuart for their unfaltering love and support


—Rolph E. Anderson, Philadelphia, PA

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Brief contents

1 Overview of Multivariate Methods 1

Section i Preparing for Multivariate Analysis 43


2 Examining Your Data 45

Section ii interdependence techniques 119


3 Exploratory Factor Analysis 121
4 Cluster Analysis 189

Section iii Dependence techniques – Metric outcomes 257


5 Multiple Regression Analysis 259
6 MANOVA: Extending ANOVA 371

Section iV Dependence techniques – non-metric outcomes 469


7 Multiple Discriminant Analysis 471
8 Logistic Regression: Regression with a Binary Dependent Variable 548

Section V Moving Beyond the Basics 601


9 Structural Equation Modeling: An Introduction 603
10 SEM: Confirmatory Factor Analysis 658
11 Testing Structural Equation Models 699
12 Advanced SEM Topics 726
13 Partial Least Squares Structural Equation Modeling (PLS-SEM) 759

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Contents

Preface xiv Recognize That Sample Size Affects All Results 30


Acknowledgments xvii Know Your Data 30
Strive for Model Parsimony 31
Look at Your Errors 31
1 overview of Multivariate Methods 1 Simplify Your Models By Separation 31
What is Multivariate Analysis? 3 Validate Your Results 32

three converging trends 4 A Structured Approach to Multivariate Model


Topic 1: Rise of Big Data 4 Building 32
Topic 2: Statistical Versus Data Mining Models 7 Stage 1: Define the Research Problem, Objectives,
Topic 3: Causal Inference 9 and Multivariate Technique to Be Used 33
Summary 9 Stage 2: Develop the Analysis Plan 33
Stage 3: Evaluate the Assumptions Underlying the
Multivariate Analysis in Statistical terms 9
Multivariate Technique 33
Some Basic concepts of Multivariate Analysis 10 Stage 4: Estimate the Multivariate Model and Assess
The Variate 10 Overall Model Fit 34
Measurement Scales 11 Stage 5: Interpret the Variate(s) 34
Measurement Error and Multivariate Measurement 13 Stage 6: Validate the Multivariate Model 34
Managing the Multivariate Model 14 A Decision Flowchart 34
Managing the Variate 14 Databases 34
Managing the Dependence Model 17 Primary Database 35
Statistical Significance Versus Statistical Power 18 Other Databases 37
Review 20
organization of the Remaining chapters 37
A classification of Multivariate techniques 21 Section I: Preparing for a Multivariate Analysis 37
Dependence Techniques 21 Section II: Interdependence Techniques 38
Interdependence Techniques 25 Sections III and IV: Dependence Techniques 38
types of Multivariate techniques 25 Section V: Moving Beyond the Basics 38
Exploratory Factor Analysis: Principal Components Online Resources: Additional Chapters 38
and Common Factor Analysis 25 Summary 39
Cluster Analysis 26
Questions 41
Multiple Regression 26
Multivariate Analysis of Variance and Covariance 26 Suggested Readings and online Resources 41
Multiple Discriminant Analysis 26 References 41
Logistic Regression 27
Structural Equation Modeling and Confirmatory Factor
Analysis 27
Partial Least Squares Structural Equation Modeling 28 Section i
Canonical Correlation 28 Preparing for Multivariate
Conjoint Analysis 28
Perceptual Mapping 29 Analysis 43
Correspondence Analysis 29
Guidelines for Multivariate Analyses and 2 examining Your Data 45
interpretation 29
introduction 49
Establish Practical Significance as Well as Statistical
Significance 30 the challenge of Big Data Research efforts 49
vi

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Contents vii

Data Management 50 Suggested Readings and online Resources 116


Data Quality 50 References 116
Summary 51
Preliminary examination of the Data 51
Univariate Profiling: Examining the Shape of the
Distribution 51
Bivariate Profiling: Examining the Relationship Between Section ii
Variables 52 interdependence techniques 119
Bivariate Profiling: Examining Group Differences 53
Multivariate Profiles 54
New Measures of Association 55 3 exploratory Factor Analysis 121
Summary 55
What is exploratory Factor Analysis? 124
Missing Data 56 A Hypothetical example of exploratory Factor
The Impact of Missing Data 56 Analysis 126
Recent Developments in Missing Data Analysis 57
Factor Analysis Decision Process 127
A Simple Example of a Missing Data Analysis 57
A Four-Step Process for Identifying Missing Data Stage 1: objectives of Factor Analysis 127
and Applying Remedies 58 Specifying the Unit of Analysis 127
An Illustration of Missing Data Diagnosis with the Achieving Data Summarization Versus Data
Four-Step Process 72 Reduction 129
Variable Selection 131
outliers 85
Using Factor Analysis with Other Multivariate
Two Different Contexts for Defining Outliers 85
Techniques 131
Impacts of Outliers 86
Classifying Outliers 87 Stage 2: Designing an exploratory Factor
Detecting and Handling Outliers 88 Analysis 132
An Illustrative Example of Analyzing Outliers 91 Variable Selection and Measurement Issues 132
Sample Size 132
testing the Assumptions of Multivariate
Correlations among Variables or Respondents 133
Analysis 93
Assessing Individual Variables Versus the Variate 93 Stage 3: Assumptions in exploratory Factor
Four Important Statistical Assumptions 94 Analysis 135
Conceptual Issues 135
Data transformations 100
Statistical Issues 135
Transformations Related to Statistical Properties 101
Summary 136
Transformations Related to Interpretation 101
Stage 4: Deriving Factors and Assessing overall
Transformations Related to Specific Relationship
Fit 136
Types 102
Selecting the Factor Extraction Method 138
Transformations Related to Simplification 103
Stopping Rules: Criteria for the Number of Factors to
General Guidelines for Transformations 104
Extract 140
An illustration of testing the Assumptions Alternatives to Principal Components and Common Factor
Underlying Multivariate Analysis 105 Analysis 144
Normality 105
Stage 5: interpreting the Factors 146
Homoscedasticity 108
The Three Processes of Factor Interpretation 146
Linearity 108
Factor Extraction 147
Summary 112
Rotation of Factors 147
incorporating nonmetric Data with Dummy Judging the Significance of Factor Loadings 151
Variables 112 Interpreting a Factor Matrix 153
Concept of Dummy Variables 112
Stage 6: Validation of exploratory Factor
Dummy Variable Coding 113
Analysis 158
Using Dummy Variables 113
Use of Replication or a Confirmatory Perspective 158
Summary 114 Assessing Factor Structure Stability 159
Questions 115 Detecting Influential Observations 159

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
viii Contents

Stage 7: Data Reduction—Additional Uses of A Managerial Overview of the Clustering Process 252
exploratory Factor Analysis Results 159 Summary 253
Selecting Surrogate Variables for Subsequent
Questions 254
Analysis 160
Creating Summated Scales 160 Suggested Readings and online Resources 255
Computing Factor Scores 163 References 255
Selecting among the Three Methods 164
An illustrative example 165
Stage 1: Objectives of Factor Analysis 165
Stage 2: Designing a Factor Analysis 165 Section iii
Stage 3: Assumptions in Factor Analysis 165
Dependence techniques – Metric
Principal Component Factor Analysis: Stages 4–7 168
Common Factor Analysis: Stages 4 and 5 181 outcomes 257
A Managerial Overview of the Results 183
Summary 184 5 Multiple Regression Analysis 259
Questions 187
What is Multiple Regression Analysis? 265
Suggested Readings and online Resources 187
Multiple Regression in the era of Big Data 265
References 187
An example of Simple and Multiple
Regression 266
Prediction Using a Single Independent Variable:
4 cluster Analysis 189 Simple Regression 267
What is cluster Analysis? 192 Prediction Using Several Independent Variables:
Cluster Analysis as a Multivariate Technique 192 Multiple Regression 269
Conceptual Development with Cluster Analysis 192 Summary 271
Necessity of Conceptual Support in Cluster Analysis 193 A Decision Process for Multiple Regression
How Does cluster Analysis Work? 193 Analysis 272
A Simple Example 194 Stage 1: objectives of Multiple Regression 273
Objective Versus Subjective Considerations 199 Research Problems Appropriate for Multiple
cluster Analysis Decision Process 199 Regression 273
Stage 1: Objectives of Cluster Analysis 199 Specifying a Statistical Relationship 274
Stage 2: Research Design in Cluster Analysis 202 Selection of Dependent and Independent Variables 275
Stage 3: Assumptions in Cluster Analysis 211 Stage 2: Research Design of a Multiple Regression
Stage 4: Deriving Clusters and Assessing Overall Fit 212 Analysis 278
Stage 5: Interpretation of the Clusters 227 Sample Size 278
Stage 6: Validation and Profiling of the Clusters 228 Creating Additional Variables 281
implication of Big Data Analytics 230 Overview 286
Challenges 230 Stage 3: Assumptions in Multiple Regression
An illustrative example 230 Analysis 287
Stage 1: Objectives of the Cluster Analysis 231 Assessing Individual Variables Versus the Variate 287
Stage 2: Research Design of the Cluster Analysis 232 Methods of Diagnosis 288
Stage 3: Assumptions in Cluster Analysis 235 Linearity of the Phenomenon 288
Stages 4–6: Employing Hierarchical and Nonhierarchical Constant Variance of the Error Term 290
Methods 235 Normality of the Error Term Distribution 291
Part 1: Hierarchical Cluster Analysis (Stage 4) 235 Independence of the Error Terms 291
Part 2: Nonhierarchical Cluster Analysis Summary 292
(Stages 4–6) 245 Stage 4: estimating the Regression Model
Examining an Alternative Cluster Solution: and Assessing overall Model Fit 292
Stages 4–6 251 Managing the Variate 292

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Contents ix

Variable Specification 294 Multivariate Procedures for Assessing Group


Variable Selection 295 Differences 377
Testing the Regression Variate for Meeting the Regression A Hypothetical illustration of MAnoVA 381
Assumptions 298 Analysis Design 381
Examining the Statistical Significance of Our Model 299 Differences from Discriminant Analysis 381
Understanding Influential Observations 302 Forming the Variate and Assessing Differences 382
Stage 5: interpreting the Regression Variate 308 A Decision Process for MAnoVA 383
Using the Regression Coefficients 308
Stage 1: objectives of MAnoVA 385
Assessing Multicollinearity 311
When Should We Use MANOVA? 385
Relative Importance of Independent Variables 317
Types of Multivariate Questions Suitable for
Summary 320
MANOVA 385
Stage 6: Validation of the Results 321 Selecting the Dependent Measures 386
Additional or Split Samples 321
Stage 2: issues in the Research Design of
Calculating the PRESS Statistic 321
MAnoVA 387
Comparing Regression Models 322
Types of Research Approaches 387
Forecasting with the Model 322
Types of Variables in Experimental Research 389
extending Multiple Regression 322 Sample Size Requirements—Overall and by
Multilevel Models 323 Group 391
Panel Models 328 Factorial Designs—Two or More Treatments 391
illustration of a Regression Analysis 331 Using Covariates—ANCOVA and MANCOVA 394
Stage 1: Objectives of Multiple Regression 331 Modeling Other Relationships Between Treatment and
Stage 2: Research Design of a Multiple Regression Outcome 396
Analysis 331 MANOVA Counterparts of Other ANOVA Designs 397
Stage 3: Assumptions in Multiple Regression A Special Case of MANOVA: Repeated Measures 397
Analysis 332 Stage 3: Assumptions of AnoVA and
Stage 4: Estimating the Regression Model and Assessing MAnoVA 398
Overall Model Fit 332 Independence 399
Stage 5: Interpreting the Regression Variate 348 Equality of Variance–Covariance Matrices 399
Stage 6: Validating the Results 353 Normality 400
evaluating Alternative Regression Models 355 Linearity and Multicollinearity among the Dependent
Confirmatory Regression Model 355 Variables 401
Use of Summated Scales as Remedies for Sensitivity to Outliers 401
Multicollinearity 357 Stage 4: estimation of the MAnoVA Model
Including a Nonmetric Independent Variable 361 and Assessing overall Fit 401
A Managerial Overview of the Results 361 Estimation with the General Linear Model 403
Summary 363 Measures for Significance Testing 403
Questions 366 Statistical Power of the Multivariate Tests 403
Estimating Additional Relationships: Mediation and
Suggested Readings and online Resources 367
Moderation 407
References 367
Stage 5: interpretation of the MAnoVA Results 410
Evaluating Covariates 410
Assessing Effects on the Dependent Variate 411
6 MAnoVA: extending AnoVA 371 Identifying Differences Between Individual Groups 415
Assessing Significance for Individual Outcome
Re-emergence of experimentation 376 Variables 417
experimental Approaches Versus other Multivariate Interpreting Mediation and Moderation 419
Methods 376 Stage 6: Validation of the Results 421
MAnoVA: extending Univariate Methods for Advanced issues: causal inference in
Assessing Group Differences 377 nonrandomized Situations 421

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
x Contents

Causality in the Social and Behavioral Sciences 422


The Potential Outcomes Approach 423
Section iV
Counterfactuals in Non-experimental Research Dependence techniques –
Designs 423
non-metric outcomes 469
Propensity Score Models 424
Overview 428
7 Multiple Discriminant Analysis 471
Summary 430
What is Discriminant Analysis? 474
illustration of a MAnoVA Analysis 430
The Variate 474
Research Setting 430
Testing Hypotheses 475
example 1: Difference Between two independent
Similarities to other Multivariate techniques 476
Groups 432
Stage 1: Objectives of the Analysis 432 Hypothetical example of Discriminant Analysis 476
Stage 2: Research Design of the MANOVA 433 A Two-Group Discriminant Analysis: Purchasers Versus
Stage 3: Assumptions in MANOVA 433 Non-purchasers 476
Stage 4: Estimation of the MANOVA Model and Assessing A Three-Group Example of Discriminant Analysis:
Overall Fit 434 Switching Intentions 481
Stage 5: Interpretation of the Results 437 the Decision Process for Discriminant Analysis 484
Summary 438 Stage 1: objectives of Discriminant Analysis 484
example 2: Difference Between K independent Descriptive Profile Analysis 485
Groups 438 Classification Purposes 485
Stage 1: Objectives of the MANOVA 438 Stage 2: Research Design for Discriminant
Stage 2: Research Design of MANOVA 439 Analysis 485
Stage 3: Assumptions IN MANOVA 439 Selecting Dependent and Independent Variables 485
Stage 4: Estimation of the MANOVA Model and Assessing Sample Size 487
Overall Fit 440 Division of the Sample 488
Stage 5: Interpretation of the Results 443
Stage 3: Assumptions of Discriminant Analysis 488
Summary 444
Impacts on Estimation and Classification 489
example 3: A Factorial Design for MAnoVA with Impacts on Interpretation 489
two independent Variables 444 Stage 4: estimation of the Discriminant Model
Stage 1: Objectives of the MANOVA 445 and Assessing overall Fit 490
Stage 2: Research Design of the MANOVA 445 Selecting an Estimation Method 491
Stage 3: Assumptions in MANOVA 447 Statistical Significance 492
Stage 4: Estimation of the MANOVA Model and Assessing Assessing Overall Model Fit 493
Overall Fit 448 Casewise Diagnostics 501
Stage 5: Interpretation of the Results 451
Stage 5: interpretation of the Results 503
Summary 452
Discriminant Weights 503
example 4: Moderation and Mediation 452 Discriminant Loadings 503
Moderation of Distribution System (X5) by Firm Partial F Values 504
Size (X3) 453 Interpretation of Two or More Functions 504
Summary 456 Which Interpretive Method to Use? 506
Mediation of Distribution System (X5) By Purchase
Stage 6: Validation of the Results 506
Level (X22) 457
Validation Procedures 506
Summary 459
Profiling Group Differences 507
A Managerial overview of the Results 459
A two-Group illustrative example 508
Summary 460 Stage 1: Objectives of Discriminant Analysis 508
Questions 463 Stage 2: Research Design for Discriminant Analysis 508
Stage 3: Assumptions of Discriminant Analysis 509
Suggested Readings and online Resources 464
Stage 4: Estimation of the Discriminant Model and
References 464 Assessing Overall Fit 509

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Contents xi

Stage 5: Interpretation of the Results 520 Stage 6: Validation of the Results 579
Stage 6: Validation of the Results 522 An illustrative example of Logistic Regression 580
A Managerial Overview 523 Stage 1: Objectives of Logistic Regression 580
A three-Group illustrative example 523 Stage 2: Research Design for Logistic Regression 580
Stage 1: Objectives of Discriminant Analysis 524 Stage 3: Assumptions of Logistic Regression 581
Stage 2: Research Design for Discriminant Stage 4: Estimation of the Logistic Regression Model and
Analysis 524 Assessing Overall Fit 581
Stage 3: Assumptions of Discriminant Analysis 524 Stage 5: Interpretation of Results 592
Stage 4: Estimation of the Discriminant Model and Stage 6: Validation of the Results 596
Assessing Overall Fit 525 A Managerial Overview 596
Stage 5: Interpretation of Three-Group Discriminant Summary 596
Analysis Results 537
Questions 598
Stage 6: Validation of the Discriminant Results 542
A Managerial Overview 543 Suggested Readings and online Resources 598

Summary 544 References 598

Questions 546
Suggested Readings and online Resources 547
References 547
Section V
8 Logistic Regression: Regression Moving Beyond the Basics 601
with a Binary Dependent
Variable 548 9 Structural equation Modeling:
What is Logistic Regression? 551 An introduction 603
the Decision Process for Logistic Regression 552 What is Structural equation Modeling? 607
Stage 1: objectives of Logistic Regression 552 Estimation of Multiple Interrelated Dependence
Explanation 552 Relationships 607
Classification 553 Incorporating Latent Variables Not Measured
Directly 608
Stage 2: Research Design for Logistic
Defining a Model 610
Regression 553
Representation of the Binary Dependent Variable 553 SeM and other Multivariate techniques 613
Sample Size 555 Similarity to Dependence Techniques 613
Use of Aggregated Data 556 Similarity to Interdependence Techniques 613
The Emergence of SEM 614
Stage 3: Assumptions of Logistic Regression 556
the Role of theory in Structural equation
Stage 4: estimation of the Logistic Regression
Modeling 614
Model and Assessing overall Fit 557
Specifying Relationships 614
Estimating the Logistic Regression Model 558
Establishing Causation 615
Assessing the Goodness-of-Fit of the Estimated
Developing a Modeling Strategy 618
Model 563
Overview of Assessing Model Fit 571 A Simple example of SeM 619
Casewise Diagnostics 571 Theory 619
Summary 572 Setting Up the Structural Equation Model for Path
Analysis 620
Stage 5: interpretation of the Results 572
The Basics of SEM Estimation and Assessment 621
Testing for Significance of the Coefficients 573
Interpreting the Coefficients 574 Six Stages in Structural equation Modeling 625
Calculating Probabilities for a Specific Value of Stage 1: Defining individual constructs 627
the Independent Variable 578 Operationalizing the Construct 627
Overview of Interpreting Coefficients 579 Pretesting 627

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
xii Contents

Stage 2: Developing and Specifying the Stage 3: Designing a Study to Produce empirical
Measurement Model 627 Results 670
SEM Notation 628 Measurement Scales in CFA 670
Creating the Measurement Model 629 SEM and Sampling 670
Stage 3: Designing a Study to Produce empirical Specifying the Model 670
Results 629 Issues in Identification 671
Issues in Research Design 629 Problems in Estimation 673
Issues in Model Estimation 633 Stage 4: Assessing Measurement Model
Stage 4: Assessing Measurement Model Validity 673
Validity 635 Assessing Fit 674
The Basics of Goodness-of-Fit 635 Path Estimates 674
Absolute Fit Indices 636 Construct Validity 675
Incremental Fit Indices 638 Model Diagnostics 677
Parsimony Fit Indices 639 Summary Example 681
Problems Associated with Using Fit Indices 639 cFA illustration 681
Unacceptable Model Specification to Achieve Fit 641 Stage 1: Defining Individual Constructs 682
Guidelines for Establishing Acceptable Stage 2: Developing the Overall Measurement
and Unacceptable Fit 641 Model 682
Stage 5: Specifying the Structural Model 643 Stage 3: Designing a Study to Produce Empirical
Results 684
Stage 6: Assessing the Structural Model
Stage 4: Assessing Measurement Model Validity 685
Validity 644
HBAT CFA Summary 692
Competitive Fit 645
CFA Results Detect Problems 693
Testing Structural Relationships 647
Summary 696
Summary 648
Questions 697
Questions 649
Suggested Readings and online Resources 697
Suggested Readings and online Resources 649
References 697
Appendix 9A: estimating Relationships Using Path
Analysis 650
Appendix 9B: SeM Abbreviations 653 11 testing Structural
Appendix 9c: Detail on Selected GoF indices 654 equation Models 699
References 656 What is a Structural Model? 700
A Simple example of a Structural Model 701
10 SeM: confirmatory Factor An overview of theory testing with SeM 702
Analysis 658 Stages in testing Structural theory 703
One-Step Versus Two-Step Approaches 703
What is confirmatory Factor Analysis? 660
Stage 5: Specifying the Structural Model 703
CFA and Exploratory Factor Analysis 660
Unit of Analysis 704
Measurement Theory and Psychometrics 661
Model Specification Using a Path Diagram 704
A Simple Example of CFA and SEM 661
Designing the Study 708
A Visual Diagram 661
Stage 6: Assessing the Structural Model Validity 710
SeM Stages for testing Measurement theory
Understanding Structural Model Fit from CFA Fit 710
Validation with cFA 663
Examine the Model Diagnostics 712
Stage 1: Defining individual constructs 663
SeM illustration 713
Stage 2: Developing the overall Measurement Stage 5: Specifying the Structural Model 713
Model 663 Stage 6: Assessing the Structural Model Validity 715
Unidimensionality 664
Summary 722
Congeneric Measurement Model 665
Items per Construct 665 Questions 723
Reflective Versus Formative Measurement 668 Suggested Readings and online Resources 723

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Contents xiii

Appendix 11A 724 Stage 1: Defining Research objectives and


References 725 Selecting constructs 768
Stage 2: Designing a Study to Produce empirical
Results 769
12 Advanced SeM topics 726 Metric Versus Nonmetric Data and Multivariate
Normality 769
Reflective Versus Formative Scales 728 Missing Data 770
Reflective Versus Formative Measurement Theory 728 Statistical Power 770
Operationalizing a Formative Measure 729 Model Complexity and Sample Size 770
Differences Between Reflective and Formative
Stage 3: Specifying the Measurement and
Measures 730
Structural Models 771
Which to Use—Reflective or Formative? 732
Measurement Theory and Models 773
Higher-order Factor Models 732 Structural Theory and Path Models 774
Empirical Concerns 733
Stage 4: Assessing Measurement Model
Theoretical Concerns 734
Validity 774
Using Second-Order Measurement Theories 735
Assessing Reflective Measurement Models 775
When to Use Higher-Order Factor Analysis 736
Assessing Formative Measurement Models 776
Multiple Groups Analysis 736 Summary 779
Measurement Model Comparisons 737
Stage 5: Assessing the Structural Model 779
Structural Model Comparisons 741
Collinearity among Predictor Constructs 779
Measurement type Bias 742 Examining the Coefficient of Determination 780
Model Specification 742 Effect Size 780
Model Interpretation 744 Blindfolding 780
Relationship types: Mediation and Moderation 744 Size and Significance of Path Coefficients 780
Mediation 745 Summary 781
Moderation 748 Stage 6: Advanced Analyses Using PLS-SeM 782
Developments in Advanced SeM Approaches 752 Multi-Group Analysis of Observed Heterogeneity 782
Longitudinal Data 752 Detecting Unobserved Heterogeneity 782
Latent Growth Models 752 Confirmatory Tetrad Analysis 782
Bayesian SEM 753 Mediation Effects 782
Summary 755 Moderation 783
Higher-Order Measurement Models 783
Questions 756
Summary 783
Suggested Readings and online Resources 757
PLS-SeM illustration 783
References 757 Theoretical PLS-SEM Path Model 784
Stage 4: Assessing Measurement Model Reliability
13 Partial Least Squares Structural and Validity 785
Path Coefficients 785
equation Modeling (PLS-SeM) 759
Construct Reliability 786
What is PLS-SeM? 764 Construct Validity 787
Structural Model 764 HBAT CCA Summary 790
Measurement Model 764 Stage 5: Assessing the Structural Model 790
Theory and Path Models in PLS-SEM 765 HBAT PLS-SEM Summary 791
The Emergence of SEM 765
Summary 792
Role of PLS-SEM Versus CB-SEM 766
Questions 793
estimation of Path Models with PLS-SeM 766
Measurement Model Estimation 766 Suggested Readings and online Resources 793
Structural Model Estimation 767 References 793
Estimating the Path Model Using the PLS-SEM
Algorithm 767
PLS-SeM Decision Process 768 Index 800

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Preface

i n more than four decades since the first edition of Multivariate Data Analysis, the fields of multivariate statistics,
and analytics in general, have evolved dramatically in several different directions for both academic and applied
researchers. In the methodological domain, we have seen a continued evolution of the more “traditional” statistical
methods such as multiple regression, ANOVA/MANOVA and exploratory factor analysis. These methods have
been extended not only in their capabilities (e.g., additional options for variable selection, measures of variable
importance, etc.), but also in their application (e.g., multi-level models and causal inference techniques). These
“traditional” methods have been augmented by a new generation of techniques represented by structural equation
modeling and partial least squares. These methods integrate many of the former methods (e.g., multiple regres-
sion and exploratory factor analysis) into new analytical techniques to perform confirmatory factor analysis and
structural model estimation. But perhaps most exciting has been the integration of methods from the fields of data
mining, machine learning and neural networks. These fields of study have remained separate for too long, repre-
senting different “cultures” in terms of approaches to data analysis. But as we discuss in Chapter 1 and throughout
the text, these two fields provide complementary approaches, each of which has advantages and disadvantages. We
hope that by acknowledging these complementarities we can in some small way increase the rate of integration
between the two fields.
The development of these analytical methods has also been greatly facilitated by the tremendous increase in
computing power available in so many formats and platforms. Today the processing power is essentially unlim-
ited as larger and larger types of problems are being tackled. The availability of these techniques has also been
expanded not only through the continued development of the traditional software packages such as SAS and it’s
counterpart JMP, IBM SPSS and STATA, as well as SmartPLS for PLS-SEM, but also the recent wide-spread use of
free, open-source, software, typified by the R-project, which has been around as far back as 1992, with roots at Bell
Labs previous to that time. Today researchers have at their disposal the widest range of software alternatives ever
available.
But perhaps the most interesting and exciting development has occurred in the past decade with the emergence
of “Big Data” and the acceptance of data-driven decisionmaking. Big Data has revolutionized the type and scope of
analyses that are now being performed into topics and areas never before imagined. The widespread availability of
both consumer-level, firm-level and event-level data has empowered researchers in both the academic and applied
domains to address questions that only a few short years ago were not even conceptualized. An accompanying trend
has been the acceptance of analytical approaches to decisionmaking at all levels. In some instances researchers had
little choice since the speed and scope of the activities (e.g., many digital and ecommerce decisions) required an
automated solution. But in other areas the widespread availability of heretofore unavailable data sources and unlim-
ited processing capacity quickly made the analytical option the first choice.
The first seven editions of this text and this latest edition have all attempted to reflect these changes within the
analytics landscape. As with our prior editions we still focus in the traditional statistical methods with an emphasis
on design, estimation and interpretation. We continually strive to reduce our reliance on statistical notation and
terminology and instead to identify the fundamental concepts which affect application of these techniques and then
express them in simple terms—the result being an applications-oriented introduction to multivariate analysis for
the non-statistician. Our commitment remains to provide a firm understanding of the statistical and managerial
principles underlying multivariate analysis so as to develop a “comfort zone” not only for the statistical but also the
practical issues involved.

xiv

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Preface xv

new Features
But the emergence of “Big Data, ” increased usage of data mining and machine learning techniques, and the accep-
tance of data-driven decision-making, has motivated us to try and provide a broader perspective on analytics than
in the past. Our goal is to recognize these three emerging trends and address how they impact the analytical domain
for both academicians and applied researchers. The eighth edition of Multivariate Data Analysis provides an updated
perspective on data analysis of all types of data as well as introducing some new perspectives and techniques that are
foundational in today’s world of analytics:
● New chapter on partial least squares structural equation modeling (PLS-SEM), an emerging technique with
equal applicability for researchers in the academic and organizational domains.
● Integration of the implications of Big Data into each of the chapters, providing some understanding of the role
of multivariate data analysis in this new era of analytics.
● Extended discussions of emerging topics, including causal treatments/inference (i.e., causal analysis of
non-experimental data as well as discussion of propensity score models) along with multi-level and panel
data models (extending regression into new research areas and providing a framework for cross-sectional/
time-series analysis).
● Updates in each chapter of technical improvements (e.g., multiple imputation for missing data treatments) as
well as the merging of basic principles from the fields of data mining and its applications.
● In addition to the new PLS-SEM chapter, the chapters on SEM have greater emphasis on psychometrics and
scale development, updated discussions on the use of reflective versus formative scaling, describe an alternative
approach for handing interactions (orthogonal moderators), as well as in-depth discussion of higher-order CFA
models, multi-group analyses, an introduction to Bayesian SEM, and an updated discussion of software avail-
ability. The multi-group discussion also includes an alternative to partial metric invariance when cross-group
variance problems are small. Additionally, an added set of variables is included with HBATSEM as a way of
illustrating diagnostic issues arising when the psychometric properties of a scale do not adhere to the rules of
thumb for measuring a latent factor. The expanded data set is available as HBATSEM6CON.
● Online resources for researchers including continued coverage from past editions of all of the analyses from the
latest versions of SAS, SPSS and SmartPLS (commands and outputs)
With the addition of the new chapter on PLS-SEM and other new content in each chapter, several chapters from
prior editions were omitted from the current edition, but are still available on the website (www.mvstats.com).
The chapters for conjoint analysis, multidimensional scaling and correspondence analysis have been formatted
in “publication ready” formats and are available to all adopters as well as interested researchers on the website.
Special thanks are due to Pei-ju Lucy Ting and Hsin-Ju Stephanie Tsai, both from University of Manchester, for
their work on revising the chapter on canonical correlation analysis in our prior edition. They updated this chap-
ter with an example using the HBAT database, added recently published material, and reorganized it to facilitate
understanding. This is one of the chapters now available on our Web site for those who wish to learn more about
this technique.
Each of these changes, and others not mentioned, will assist readers in gaining a more thorough understanding
of both the statistical and applied issues underlying these techniques.

PeDAGoGY
Almost all statistics texts include formulas and require knowledge of calculus or matrix algebra. A very important
question is “Can students comprehend what they are reading and apply it?” This book offers a wealth of pedagogical
features, all aimed at answering this question positively. Presented here is a list of the major elements:
Learning Objectives. Each chapter begins with clear learning objectives that students can use to assess their
expectations for the chapter in view of the nature and importance of the chapter material.

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
xvi Preface

Key Terms and Concepts. These are bold-faced in the text and are listed at the beginning of the chapters to facil-
itate comprehension.
Chapter Summaries. These detailed summaries are organized by the learning objectives presented at the
beginning of the chapter. This approach to organizing summaries helps students to remember the key facts,
concepts, and issues. They also serve as an excellent study guide to prepare for in-class exercises or exams.
Questions for Review and Discussion. The review and discussion questions are carefully designed to enhance the
self-learning process and to encourage application of the concepts learned in the chapter. There are six or seven
questions in each chapter designed to provide students with opportunities to enhance their comprehension of
the concepts.
Suggested Readings. A list of the most relevant additional readings is provided at the end of the chapter. These
readings enable you to review many of the sources of the information summarized in the chapter as well as extend
your knowledge to other more detailed information.
HBAT Database. The HBAT database is a continuing case of a business scenario embedded throughout the book
for the purpose of illustrating the various statistical concepts and methods. The case is introduced in Chapter 1,
and in each subsequent chapter it builds upon the previously learned concepts. A single research situation is used
to illustrate various aspects of the process of applying each of the multivariate methods. The HBATSEM data is
enhanced for the 8th edition.
Software commands/syntax. An expanded feature from prior editions are the software-specific resources to
enable a researcher to replicate the analyses in the text in SAS, IBM SPSS, and SmartPLS for that chapter. In addi-
tion to these software commands, actual outputs for all of the analyses in the text are available for the student to
examine and even compare to their results. The authors are also committed to continue development of these
resources, hopefully extending theses supplements to include selected R code and outputs in the near future.

onLine ReSoURceS
The book provides an extensive and rich ancillary package. The following is a brief description of each element in
the package. These materials are available via Cengage Brain and the text’s dedicated website (www.mvstats.com).
Instructor’s Resources. PowerPoint slide presentations provide an easy transition for instructors teaching with
the book the first time. For those who have used previous editions, there are many new support materials to build
upon the notes and teaching enhancement materials available previously. A wealth of extra student projects and
examples are available as additional classroom resources. All of these materials are available via Cengage Brain.
Website. Students can access the book’s dedicated website (www.mvstats.com) for additional information about
the various statistical methods and to evaluate their understanding of chapter material. Additional resources are
offered for each chapter—look for prompts in the book that will guide you to the website for more useful infor-
mation on various topics.
Data Sets. Data sets in SPSS format are available at the book’s website (www.mvstats.com). The two primary
data sets are the HBAT customer survey data set used in the first 8 chapters of the book, and the HBAT employee
survey data set used in Chapters 9 to 13. These data sets can be used to illustrate the examples in the text as well
as assign application exercises that go beyond the book examples.

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Acknowledgments

W e would like to acknowledge the comments and suggestions by Dr. Stephen Vaisey of Duke University,
Dr. Helmut Schneider of Louisiana State University, and Dr. Marko Sarstedt, Otto-von-Guericke-University,
Germany, on this edition. We would also like to acknowledge the assistance of the following individuals on prior
editions of the text: Bruce Alford, Louisiana Tech University; David Andrus, Kansas State University; Jill Attaway,
Illinois State University; David Booth, Kent State University; Jim Boles, University of North Carolina-Greensboro;
Alvin C. Burns, Louisiana State University; Alan J. Bush, University of Memphis; Robert Bush, Louisiana State
University at Alexandria; Rabikar Chatterjee, University of Michigan; Kerri Curtis, Golden Gate University; Chaim
Ehrman, University of Illinois at Chicago; Joel Evans, Hofstra University; Thomas L. Gillpatrick, Portland State
University; Andreas Herrman, University of St. Gallen; Dipak Jain, Northwestern University; Stavros Kalafatis,
Kingston University; John Lastovicka, University of Kansas; Margaret Liebman, La Salle University; Arthur Money,
Henley Management College; Peter McGoldrick, University of Manchester; Richard Netemeyer, University of
Virginia; Ossi Pesamaa, Jonkoping University, Robert Peterson, University of Texas; Torsten Pieper, Kennesaw State
University; Scott Roach, Northeast Louisiana University; Phillip Samouel, Kingston University; Marcus Schmidt,
Copenhagen Business School; Muzaffar Shaikh, Florida Institute of Technology; Dan Sherrell, University of
Memphis; Walter A. Smith, Tulsa University; Goren Svensson, University of Oslo; Ronald D. Taylor, Mississippi State
University; Lucy Ting, University of Manchester; Arch Woodside, Boston College; and Jerry L. Wall, University of
Louisiana-Monroe. Finally, we could not have written this or previous editions without the extensive feedback and
input from our many graduate students who provided invaluable guidance on the content and direction of the book.
J.F.H.
W.C.B.
B.J.B.
R.E.A.

xvii

Copyright 2019 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Another random document with
no related content on Scribd:
II.

Lienalla ei olisi Matin reessä ollut varsin pitkälti kuljettavaa


kirkolle, mutta Matti kiusasi häntä lähtemään edelleen myllylle asti.
He olivat nuorempina usein leikkineet ja oleksineet yhdessä, vaikka
Matti olikin pari vuotta vanhempi.

"Mitäpä minulla siellä myllyllä olisi tekemistä?" vastusteli Liena.

"Siellähän kerron", tuumi Matti ja kiirehti hepoaan.

Piti kuitenkin pysähtyä suntion pihaan, että Liena ehti jättää


nyyttinsä ja sanoa, mihin menee, että kotiväki ei kaihoelisi, jos hän
vähän viivähtäisi enemmän, kuin Vaarassa käyntiin olisi aikaa
tarvinnut.

Edelleen ajaessa pysyi Matti alkumatkan harvapuheisena, tavasta


vain muistutellen hepoansa. Viimein hän yht'äkkiä virkahti:

"Kuulehan, Liena, tulisitko sinä meidän taloon miniäksi?"

"Mikolleko sitte vaiko Jussille?" ivasi Liena veitikkamaisesti


hymyillen.. "Maltahan kun sentään ajattelen. Ensinnäkin on teillä niin
kelvottoman siivotonta. Ulkona elukat oleksivat aivan rappusien
edessä, niiden nurkkia pureksien, jopa osa porstuassakin, kuten
tänään. Entäs se koko teidän pihanne! Jos minä sinne tulisin, pitäisi
ainakin kevät- ja syysajoiksi hankkia vene, jolla saisin rappusilta
läävään soutaa, ja sinä kai olisit soutajana, eikö niin? — Ja sisällä;
no, pitäisinpähän minä siellä parempaa järjestystä."

"Sinä vain kaikki leikiksi käännät. Totta minä tarkoitan. Minähän


sinusta vanhemmilleni miniän tekisin."

"Vai niin. No, pitäneehän minunkin sitte totta puhua. Kyllä minä
sinusta pidän, Matti. Vaan osaakohan sinuun oikein luottaa? Ne
siellä teillä tuntuvat niin pettelevän toisiaan. Mitenkä lieneekään
sinun vakavuutesi. Sanopas, mihin tuo pussi on menemässä!"

Liena osoitti sevältä heinäin alta pilkistävää pussin suuta.

"Senpähän kohta näet."

Samassa reki kääntyi myllypihaan.

Matti pysäytti hevosensa puodin eteen, joka oli melkein kiinni


maantiessä, ja pyysi Lienaa puotiin. Siellä hän otatti pöydälle koko
kasat huiveja, käski Lienaa niistä valitsemaan itselleen parasta,
kuiskasi puotimiehelle jotakin ja läksi puotipojan kanssa ulos, joka
näkyi jo asiat tajuavan. Pussi palasi tuota pikaa kauppiaan
makasiinista tyhjänä rekeen, ja Matti ajoi muun kuorman suoraa
päätä myllyn ovesta sisään.

Jauhattajia ei sattunutkaan olemaan muita kuin pari pussimiestä.


Matin riisuessa ja asettaessa hevostaan syömään loppuivat nekin
jauheet, joten hän pian selvisi joutilaaksi muutamaksi tunniksi.
Matti palasi puotiin, panetti eräästä laatikosta jotakin
paperitötteröön ja antoi sen Lienalle.

"Jos minun nyt välttämättä pitää ottaa jotakin, niin otanpahan sitte
tämän."

Liena näytti keskikokoista silkkihuivia.

"Onko sinulla muuta ostamista?"

"Ei itselleni, vaan äiti se käski ottamaan kahvia."

Matti punnitutti kaksi naulaa puolekkain ja Lienan yrittäessä


maksamaan iski silmää puotimiehelle, joka kohta selitti Lienalle, että
ei se mitään maksa.

Liena tiesi seudun tavat eikä ollut niin millänsäkään, sanoipahan


vain leikkisästi:

"Hyvinhän nyt myllypuoti on rikastunut, kun jo ilmaiseksi jakelee


tavaroitaan."

Matti läksi Lienaa saattamaan kotimatkalle jalkaisin. Jo pihalle


päästyä kysäsi Liena:

"Tähänkö se elopussi nyt hupeni?"

"Huvan pienihän se oli hupenemattakin."

"Kuulepas, Matti, tuskin vain minä sinuun uskallan luottaa.


Pettelethän sinäkin kotiväkeäsi."

"Harvoin se on tapahtunut enkä nytkään sitä olisi muuten tehnyt,


vaan näin sinun tulevan. Mitenpä täällä muuten minkään talon poika
saisi rahaa? Työtä kyllä saa tehdä kuin pahin renki, mutta älä
menekään isältä pyytämään, hän tarvitsee rahat muuhun. Jo olen
välistä arvellut lähteä muualle rengiksi, saisinhan silloin edes hyvän
palkan."

"On tuossa totta, mitä puhut. Vaan tosiaan, minä en voi tulla teille
miniäksi, niin kuin asiat nyt ovat. Sinä kun olet järkevä mies, puhu
sinä suoraan isäsi kanssa ja hän puhukoon äidillesi. Minä kyllä sitte
koetan pitää Anua Maijasta huolta.

"Onpa se Mikkokin jo aika velikulta.

"Hänet voitte varsin hyvin panna kouluun ja samoin Jussin, kuin


siksi varttuu.

"Jopa se Mikko alkaa olla liian suuri mies kouluun pantavaksi.


Jospa hän siellä vain oppii enemmän pahuutta, kun ei ole ketään
silmällä pitämässä. Kotona edes joskus saa kuria."

"Ole huoletta, kyllä siellä koulussa tehdään kaikki ihmisiksi",


vakuutti Liena, uskoen opetukseen pystyvän kelvottomiin yhtä
helposti kuin häneen itseensä, jolla oli ollut kunnon vanhemmat
kasvattajina.

Matti lupasi koettaa parastansa, vaan eivätpä siinä kaikki


vastukset olleet. Häntä itseäänkin oli ainiaan harmittanut
siivottomuus, mutta mitä sille kenkään voi tehdä. Karjopiha oli
luonnostaan niin vesinen, ja mitenpä niitä elukoita voikaan saada
pysymään poissa rappusilta. Ja kun kerran aivan rappusten edessä
oli likaa puolisääreen, niin siitähän ne likautuivat lattiatkin.
Lienalle johtui mieleen, että Vaaran vanhempi eli nykyinen
asuintupa aiottiin juuri keväällä purkaa ja että oli vedetty pitkin talvea
hirsiä rakennustarpeiksi.

"Kääntäkää koko tuparakennus toisin päin ja pankaa aita


pihamaan ja karjopihan välille."

"Tuleepa suotta maksamaan kylmän tuvan muutto."

"Eihän sitä tarvitsekaan muuttaa. Asettakaa uusi porstua ja


maitokammari kylmän tuvan toista sivuseinää vasten ja jatkakaa
kartanoa sinne päin niin tulevat rappuset pois karjopihasta."

"Ei tuo taitaisi hulluinta olla, jos vain kotiväki siihen suostuisi."

"Olisi hyvä aidata jokin kapanala pihamaaksi ja laittaa siihen pieni


kasvitarha. Siitäpä tulisikin oikein sievä laitos."

"Sinullahan nuo näkyvät kukat kotonasikin hyvin menestyvän ja


marjapensaat."

"Ja tehkää sitte uuden tuvan päähän kaksi pientä kammaria."

Mattia alkoi arveluttaa. Eiköhän vain Liena ruvennut jo vaatimaan


aivan mahdottomia.

"Jos sattuisi joku sairastumaan, olisi niin hyvä hänen olla rauhassa
muiden hälinältä."

Se naula veti. Matti oli itse ennen joulua ollut sairaana ja hän kyllä
muisti, miten lasten parku ja työn pauke olivat häntä vaivanneet.

Kotvasen he vielä siinä astuessaan rakensivat uutta tupaa


kammareineen, tasoittelivat ja täyttivät pihamaata, tekivät
kasvitarhaa ja panivat aitaa, kunnes saapuivat kirkkomäen alle, josta
Matti ei enää lähtenyt ylös astumaan.

Kätellee erottuaan Matista, jatkoi Liena iloisesti hyräillen


astuntaansa ylös pitäjäntuvalle. Iloinen hänen mielensä nyt
todellakin ole. Vähän tosin kammotti tuo Vaaran kotielämä, mutta
"tottapahan sitte yksin neuvoin selvitellään”, ajatteli hän.

"Mutta olenhan itsekin aivan samanlainen!" johtui hänelle sitte


mieleen. "Enkö ottanut vastaan salatavaraa Vaarassa ja nyt
uudestaan puodissa! Vaan vaikeahan on syrjästä päin sekautua
heidän asioihinsa. Pidänpähän vasta paremmin varani ja säilytän
omantuntoni puhtaana. Jospa vain Matti saisi lupauksensa
toteutumaan!"

Matti puolestansa riensi hyvää vauhtia takaisin myllylle ja läksi,


saatuaan elot jauhoina rekeen, hyvällä mielellä ajamaan kotiinsa.

Huudettuaan kotipihassa Mikon tuvasta korjaamaan hevosta, kun


näet ei jauhorekeä käynyt jättää ahnasten sikojen haltuun, sekä
kannettuaan säkit aittaan ja tyhjennettyään salvoon, astui Matti,
ajokalut käsivarrella, tupaan, jossa koko perhe istui iltapuhde-
töissään.

Isäntä veisteli kaplasta rekeen, joka oli leikkautunut ja seisoi


syrjällään penkin vieressä sulamassa.

Emäntä kehräsi villalepeitä vasusta, johon piika Tiina koetti niitä


kartata toisia sijaan.

Anna Maija neuloi sukkaa, Mikko koetti muka kiskoa päreitä ja


Jussi käpelehti isänsä vestolastujen parissa.
Kaikki istuivat piirissä päretulen ympärillä, joka paloi pihtiin
pistettynä uunin ulkonurkassa.

Matilla ei nyt ollut käsillä muuta työtä kuin Mikon kanssa yhteinen.
Siinä hänellä kyllä oli lomaa alkaa puhetta. Odotettuaan, kunnes
Tiina läksi karjaa illastamaan, sanoi hän äkisti:

"Kuulehan, isä, etkö rupeisi maksamaan minulle palkkaa, elossa


tai rahassa, sama se?"

Isäntä ja emäntä ällistyivät. Kotvanen kului ennen kuin isäntä


virkkoi:

"Jopa nyt hulluttelet, poika. Vai pitäisi tässä ruveta palkan


maksuun omille lapsilleen!"

"Oikein minä totta puhun. Enkä minä paljoa tahtoisikaan. Kunhan


edes vähänkään lupaisit, niin minun ei tarvitsisi omin luvin ottaa,
kuten tänäkin päivänä." Ja Matti huokasi syvään.

Isännältä pysähtyi työ. Hän luuli Matin laskevan leikkiä. Vaan


katsoessaan kyyrysistään häntä silmiin ja nähdessään tavallisen
leikin olevan niistä tällä kertaa kaukana nousi hän suoraksi istumaan
ja muuttui vakavaksi.

"Vai alat jo sinäkin niillä jäljillä kävellä? Akkaväen tehtäväähän se


toki tähän asti on ollut."

"Joko taas on akkaväki hampaissasi!" ärähti emäntä. "Mitenpä


tässä muuten työt teetettäisiin?"

Anna Maijalla ei näyttänyt olevan hyvää rauhaa. Hän oli ryhtynyt


Tiinan työtä jatkamaan, vaan siirtyi nyt punastuen edemmäksi tulen
valosta.

"Tahdotko ehkä sinäkin palkkaa? Jopa sitte pitäisi olla suuri


kukkaro."

"No, isä, koetetaan nyt selvittää asiat suuttumatta. Mikä menee


talon tarpeihin, samahan se, meneekö se rahassa vaiko muussa
tavarassa, vaikka tavarassa maksaminen taitaa sentään käydä
helpommaksi kuin rahassa. Eikä minun mielestäni tarvitsisi
salatakaan kehruuttamisia ja ompeluttamisia ja muuta sellaista, mitä
ei kotona ehditä eikä osata tehdä. Vaikka parempi olisi ollut käyttää
Anna Maijaa kansakoulussa, niin osaisihan nyt edes ommella
kotitarpeet, ja hänellähän siihen olisi aikaa. Tottapa se muukin oppi
edestä löytyisi elämän matkan varrelta."

Tämän puheen aikana olivat isännän kasvot osoittaneet milloin


tyytymättömyyttä, milloin mieltymystä. Nyt hän vastasi:

"Enhän minä ole kieltänytkään töitä teettämästä. Tuo elon ja voin


kantaminen puotiin kaikenlaisesta joutavasta räämästä, se se minua
suututtaa, kun sitä lisäksi vielä tehdään salaa ja varkain.

"Siinäpä se on juuri. Liika on tietysti aina liika, ja koreus ja ylpeys


käy lankeemuksen edellä. Mutta olisinpa tänään tarvinnut muutamia
markkoja, ja kun en rahaa kotoa saa, täytyi minun ottaa, mitä sain;
vein rukiita puotiin. Teenhän tässä työtä minäkin ja uskollisemmin
kuin mikään renki.

"No, mihinkä sinä nyt olisit niin rahaa tarvinnut? Etkö saattanut
pyytää?

"Metsässäpä olit, ja turha tuo on ennenkin ollut pyytö.


"Poika, poika, älä puolustele varkautta!" torui isä syvästi huoaten.
"Ei se sellainen menesty.

"Kylläpä sitte pian salvot tyhjenevät, jos pojat niitä rupeavat puotiin
vetämään oikein hevosella!” ihmetteli emäntä.

"Ikään kuin eivät salvoja muut puolentelisi kuin minä. — Mutta


varkaudelta se itsestäni tuntui tänään, ja hyvin koetin katsoa, että
saisin pussin salaa rekeen, vaikka toisa päin ajattelinkin, että
omaanihan siinä kannan.

"Näin minä ja näki suntion Lienakin, toimitti Anna Maija, unhottaen


omat jälkensä.

"Etkä virkkanut mitään, kun Jussi puhui pussista!” nuhteli emäntä.

"Mitenpä hän virkkoi, kun hänellä itselläänkin oli muuta mielessä.


Asialla se Liena täällä kävi."

"Kyllä tässä ihmeissä käsin ollaan! Kaikki vain taloa varastavat ja


vielä julkeavat noin ilmi puhua. Onhan tässä koetettu neuvoa ja
kurittaa; missä sitte lienee vika? Sinäkin, Matti, teet tuolla lailla,
vaikka olen sinuun aina luottanut. Etkö luule sen tuntuvan karvaalta
isäsi mielestä?"

"Isä, älähän nyt sentään minua ihan umpimähkään tuomitse!"

"Mihin sinä sitte tarvitsit rahaa? Vastaa siihen ensin ja puolustaudu


vasta sitte, jos osaat."

"Sitä en nyt tahtoisi sanoa tällä kertaa. Uskokaa, kun vakuutan,


että turhuuteen en pannut. Tuonempana kyllä kerron, ja silloin sen
itsekin hyväksytte."
"No, samahan tuo sitte oli, jos sait rahaa tai ruista; enkä minä niin
kovin moiti omin luvin otostakaan.

"Mutta minun omatuntoni moittii. Minä juuri tahtoisin päästä


salaamasta ja sen tähden minä pyysin palkkaa. Lupaa vaikka viisi-
tai vain neljäkin kymmentä markkaa. Jos elot riittää, myödään silloin,
kuin saadaan hintaa; jos eivät riitä myötäväksi, niin vähempihän
ainakin tarvitaan ostaa. Ei siitä talolle vahinkoa ole, vaan minulle
rauhallinen mieli; omaanihan saisin käyttää. Jos en kaikkea tarvitsisi,
niin talteenpahan jäisi minunkin taakseni."

Isäntä oli paikkakuntansa miehiä ja pysyi mielellään vanhoissa


tavoissa, mutta ei kuitenkaan ollut niin itsepäinen eikä
ymmärtämätön, että ei olisi taipunut järkevistä syistä: eihän hän ollut
keltään vaatinut tähänkään asti sokeaa tottelemista, vaan päin
vastoin hän olisi mielellään neuvotellut kaikista talon asioista
jokaisen kanssa, jos vain toiset olisivat siihen taipuneet. Mutta kun
varsinkin emäntä teki kaikki asiansa salaa ja ne kuitenkin aina tulivat
perästä päin tavalla tai toisella ilmi, tuli isäntä siitä usein
muistuttaneeksi. Tästä oli yksin suoravaisen Matinkin ja tietysti vielä
enemmin Anna Maijan mieleen kasvanut jonkinlaista kammoa, joka
heitä pakotti seuraamaan äitinsä esimerkkiä silloinkin, kuin ei ollut
tarviskaan salaella.

Nyt isäntä vähän aikaa mietiskeltyään virkkoi leppeästi:

"Et sinä, poika, näy tyhmimpiä olevan. Oudolta tuo tuntuu palkan
maksu omalle pojalleen, vaan parempi kuitenkin on maksaa sinulle
kuin rengille, joka saattaisi olla petollinen silmän palvelija."

"Sitähän juuri minäkin ajattelin. Ei minun nyt siis tarvitse miettiä


muuanne rengiksi lähtöä, kuten on välistä mieleeni jolahtanut."
"Ei sinun toki olisi tarvinnut edes ajatellakaan sellaisen häpeän
tekoa omalle kodillesi."

Näin oli Matin suhde sovitettu. Mutta ikään kuin korvatakseen


poikkeamistansa yleisestä tavasta kysyi isäntä vähän ivan sekaisella
äänellä, pitikö muillekin määrätä palkka.

Emäntä yritti suuttumaan ja Anna Maija jo salaa iloitsi, että ehkä


hänkin saa oman kukkaron.

Matti, tietäen sen toivon turhaksi ja katsoen itsekin niin moneen


palaan osittelemisen tarpeettomaksi, osasi selvittää asian kaikkein
mieliksi siten, että talon työt päätettiin teettää entiseen tapaan, vaan
salaamatta, ja että kukin sai välttämättömiin omiin tarpeihinsa pyytää
rahaa tai tavaraa isännältä, joka kuitenkin jo edeltä päin jyrkkään eitti
mitään antavansa mihinkään koreilemiseen, koska talon varat sitä
eivät kannattaneet.

Näin uskoivat kaikki nyt salailemisen ja petoksen olevan kotiväen


kesken kerrassaan haudatun, ja isäntä toivoi samassa suhteessa
myöskin kasvavan rehellisyyden ja suoruuden muita kohtaan.

Keskustelun päätyttyä Mikko isänsä käskystä etsi virsikirjan ja luki


iltarukouksen sekä Herran rukouksen ja siunauksen, kuten tavallista
oli Vaarassa, milloin isäntä itse ei päivää päättänyt siunauksella.

*****

Pahempi oli rakennus- ja siitä riippuva siivoasia selvitellä, jonka


Matti vasta toisena iltana otti puheeksi.

Tuvansijan muuttoa ja karjapihan aitaamista erilleen pihamaasta


ensin kaikki jyrkästi vastustivat. Kuka tuota viitsi ruveta aitain ylitse
joka kerran kapuamaan karjan asioille mennessä ja kuka muisti pitää
porttia kiinni, jos sellainen korukalu tehtäisiinkin. Ja kun viimein
ehdittiin jo vähän tottua näiden hankaluuksien ajatukseen, ilmestyi
muita vastuksia; milloin oli katto, milloin uunit esteenä.

Tätä neuvottelua kesti monta viikkoa. Sen verran Matti kuitenkin


sai aikaan, että vedettiin metsästä varalle enempi hirsiä ja kiviäkin
koottiin runsaammin.

Hauskalta nyt työ tuntui Matista. Hän raatoi kuin karhu,


vahvemmin kuin isänsä, vaikka hänkin vielä oli voimissaan. No,
Matillahan olikin mieleinen toivo tavoitettavana.

Moneen kertaan Matti mietiskeli ja jo ulkonakin lumeen


piirrustellen aprikoitsi, miten saisi kaikki kotiväen lukemattomat
estelyt kumotuksi.

Viimein hän eräänä sunnuntaina jäi kirkosta suntioon ja ryhtyi


Lienan kanssa asiaa perin pohjin harkitsemaan.

Liena oli koulussa oppinut piirrustelemaan kuvia monenkin laisia,


vaikka juuri ei rakennusten pohjapiirroksia. Se taito tuli nyt hyvään
tarpeesen. Illempana kotiin lähtiessään Matti pisti uuden
asuinkartanon pohjakuvan taskuunsa.

Siinä Lienan mestariteoksessa oli kaikki muut vastukset saatu


poistetuksi paitsi paikoilleen jäävän tuvan katon kääntäminen, joka
jäi tarpeettomaksi liikatyöksi. Oli siihen myöskin uuden tuvan jatkoksi
piirretty kaksi kammaria, jotka sentään voitiin jättää poiskin, jos
Vaaran isäntä hyvin jyrkästi pani vastaan.
Kotiin ehdittyänsä ja riisuuduttuansa levitti Matti paperin pöydälle
isänsä eteen, sanoen:

"Tuossa se nyt on uusi tuparakennus."

Siihenpä heti kaikki muutkin kokoutuivat ja Matin piti juurta jaksain


selittää kaikki tyyni, missä oli ovet, ikkunat ja uunit, rappuset ja muut
sellaiset.

"No, kylmiksikö ne nuo kammarit jätetäänkin?" kysyi emäntä.


"Kuka sairas niissä sitte voi elää?"

"Mitäpä niihin uuneista, kun ei ole tietoa seinäinkään teosta."

"Sen minä sanon, että jos kerran kartanon sijaa muutetaan, niin
kyllä ne pitää tehdä kammaritkin", päätti emäntä.

"No, no, vai sinä tässä rapeatkin rakentajaksi!" virkkoi isäntä.

Asia muuttui leikiksi ja samalla se oli päätetty.

Huomeisaamuna oli Matti vielä varemmin jalkeella. Kun isäntä


astui vaatteissaan ulos, olivat jo hevoset hirsirekien eteen valjastetut
ja Matti reessä istuen odottamassa.
III.

Seuraavana syksynä seisoi jo Vaaran tuparakennus uutuudesta


hohtavana, poikittain käännettynä ja muutenkin Lienan piirrustuksen
mukaan tehtynä, kammareinensa kaikkinensa valmiina.

Tietysti oli myöskin varsinaisia kirvesmiehiä ollut rakennustyössä,


mutta hyvää apuapa oli Matistakin ollut, milloin hän suinkin talon
töiltä jouti. Varsinkin kevätaikana oli Matti melkein nukahtamatta yöt
päivät kaivanut kivijalan sijaa veistellyt hirsiä ja tehnyt kaikkea
muuta, mihin vain yksin pystyi ja mihin hänen taitonsa riitti.

Suntion Lienalle oli hän jo kevätkorvalla tarkoilleen kertonut, miten


hänen oli ainakin aluksi onnistunut täyttää Lienan ehdot.

Kun vanha tupa lämpimämpäin ilmain tultua purettiin ja uusi


kivijalka seisoi valmiina sekä yhdet pohjahirret saatiin eräänä
lauantai-iltana kivien päälle paikoilleen, toi Matti pyhänä Lienan
Vaaraan ja kertoi kotiväelle, mitä varten hän oli rukiita puotiin vienyt,
vieläpä jäänyt vähän velkaankin, vaikka sen sittemmin jo oli kuitiksi
suorittanut.

Sattumalta, kenties salaisesta kutsustakin, kokoutui Vaaraan


muutakin nuorisoa ja suntion vanhukset, niin että voitiin vierasten
puolesta ryhtyä tupakaisten pitoon, ja ne silloin pitemmittä mutkitta ja
verukkeitta vietettiinkin nyt asuttavana olevassa ainoassa tuvassa.

Kuuliaiset jätettiin kesemmäksi, vaan usein Liena kävi Vaarassa


katsomassa, miten seinät kohoamistaan kohosivat, ja tinkimässä
kasvitarhan alaa sekä kylvämässä siksi erotetun hyvämaisen
peltokaistaleen ulkolaitaan, loitommaksi työmiesten jaloista, kukan
siemeniä.

Häät vietettiin jo uudessa tuvassa, vaikka siitä vielä silloin puuttui


uuni ja lattiakin oli vain hätävaraksi paikoilleen asetettu eikä
ikkunoita ollut hituakaan. Tarkenihan siinä kesäiltana ja yönä
tanssiessa niinkin.

Näissä häissä eräs siistipukuinen, vähän herrasmieheen vivahtava


nuorukainen hyvin ahkerasti piti tanssia vireillä, enimmäkseen,
melkeinpä yksinomaan tanssien vain talon tyttären Anna Maijan
kanssa ja hyvin ihastuneesti hänen silmiinsä katsellen.

Tämä nuorukainen oli Salonmökin Janne, joka äskettäin oli


kaupungissa päättänyt oppiaikansa ja valmiina nikkarimiehenä
palannut kotiseudulleen vähän levähtämään ja harrastelemaan.
Tosin hän oli aikonut parin viikon kuluttua lähteä suurempiin
työpaikkoihin, mutta Anna Maija hänet tänä iltana niin lumosi, että
hän vastustelematta suostui Vaaran isännän pyynnöstä jäämään
ikkunain ja ovien tekoon uuteen tupaan, vaikka hänelle siitä työstä ei
luvattukaan niin suurta palkkaa, kuin kaupunkipaikoista olisi saanut.
Niinpä hän seuraavana päivänä kävi salolta noutamassa työkalunsa
ja vanhan äitinsä ja asettui toistaiseksi riiheen asumaan ja
nikkaroimaan.
Syksymmällä ei enää muita ansiopaikkoja kuulunut
etäämmälläkään. Olihan siinä kyllin syytä pyytää Vaaran isännältä,
uuden tuvan valmistuttua, toista tupaa nikkarin työhuoneeksi ja,
isännän viimein vaikka vähän vastahakoisestikin suostuttua, asettua
siihen talvikaudeksi korjailemaa pitäjän herrasväen rikkinäisiä
huonekaluja, jopa tekemään uusiakin.

Liena, päästyään Vaaraan miniäksi, ryhtyi heti panemaan toimeen


parempaa järjestystä ensinnä maitokammarissa, jonka hoidon
emäntä mielellään jätti hänelle. Pääsihän hän siten siitä huolesta,
arveli emäntä itsekseen. Niinpä hävisivät sieltä kerrassaan kaikki
tomukasat.

Asuintuvan lattia ei vielä ollut ehtinyt varsin pahaksi mustua, ja


Liena nyt tietysti piti huolta, että se edelleen mustumaan ei
päässytkään. Tämä tosin tuntui välistä vähän liialliselta emännän
mielestä, niin että hän jo yritteli sanomaan: "mitä tuosta nyt joka
pyhäksi viitsit hangata? likautuuhan se kuitenkin." Mutta kun Liena
säveästi vastasi: "no, hyvä äiti, emmehän nyt rupea liassa elämään",
ei emäntä sentään lujemmin huolinut vastustaa, ajattelihan vain:
"kun ehtii ja viitsii, niin pesköön vaikka joka päivä."

Pitkään aikaan ei emännällä nyt ollut mitään salailemisen syitä.


Sisar Marin mies oli parannut, niin että he nyt kesäiseen aikaan
tulivat toimeen avuttakin. Isäntä toi kylliksi kahvia rukiitten sekaan,
sokuria ja muita talon tarpeita, eikä nyt myöskään enää tarvittu
teettää kylässä joka kohtaa, kuten ennen Lienan tuloa, sillä hän
osasi ja ehti itse paljon, opettipa vielä Anna Maijaakin.

Huomattuaan Lienan taitavaksi alkoi emäntä vähitellen kysellä


häneltä neuvoa, ensin tosin aristellen, että ei liiaksi luopuisi
emännän arvostaan, vaan sitte yhä useammin ja suorempaan, ja jos
Anna Maija jotakin kysyi äidiltään, neuvoi äiti häntä Lienan luo.

Kun tätä paitsi Matti aina ja usein isäntäkin neuvotteli Lienan


kanssa, kohosi hän tahtomattansa koko talon hallitsijaksi, ja hyvin
näytti kaikki käyvän, paremmin kuin Matti ja Liena olivat uskaltaneet
toivoakaan. Ei kellään ollut mitään valittamisen aihetta muusta kuin
että Mikko pyrki, milloin vain soveltui, sieppaamaan sokuripalaa tai
muuta mieluista.

Alkoipa siitä syksy lähestyä tai oikeastaan syyskuun alku ja Liena


tahtoi kaikin mokomin toimittaa Mikkoa kouluun. Hyvin isäntä ensin
vastusteli moista ajan ja tavaran tuhlausta, ei juuri itse opin tähden,
vaan varain riittämättömyyden pelosta. Mutta taipui hän sentään, kun
Liena huomautti tuota Mikossa jo näkemäänsä, isännälle
ennestäänkin tiettyä näpistelemiseen taipumusta ja lausui
toivovansa sen siellä koulussa katoavan.

Niinpä Mikko vietiin lähikaupungin kansakouluun ja pantiin evästä


mukaan muutamiksi viikoiksi sekä luvattiin kyllä ajoissa tuoda lisää.

Liianpa aikaisin Liena sentään alkoi riemuita työnsä


menestyksestä.

Nikkari Janne oli ensi hetkestä asti saavuttanut emännän suosion.


Olihan hän pulska poiki ja sitä paitsi mestarimies. Eipä siis ihme, että
emännän mielessä alkoi häilyä "nikkarin matamin" nimitys
sointuvampana kuin "talon emäntä", eikä hän malttanut olla sitä
huomauttamatta Anna Maijalle. Samaan aikaan hänestä rupesi
näyttämään Jannen äidin voipytty niin omaisen tutulta, että siihen
välistä livahti talon pytystä neljännes- tai puolinaulainen "siru".
Maitoa osui myöskin "tilkka" salatietä saman mummon haltuun. Ja
Jannen äiti puolestaan oli hyvin halukas ottamaan vastaan ilmaiseksi
annettua; siihenhän hän oli koko ikänsä tottunutkin köyhyydessä
eläessään.

Liena tätä välistä huomasi, mutta ei ryhtynyt emäntää opettamaan;


sehän toki olisi ollut sopimatonta. Matillensa hän kahden kesken
asian mainitsi ja pahoitteli, että arvattavasti Anna Maijakin eksyy
samalle tielle.

Joulun tullessa oli Matin lähdettävä noutamaan Mikkoa luvalle


kotiin luvun alottelusta.

Liena pyysi häntä tuomaan samalla kaupungista paperia, että saisi


ruveta kirjanpitoon tuloista ja menoista, jollaista hänelle ennen
koulussa oli vaikka hyvin niukastikin muutaman kerran vain
näyttämällä opetettu.

"Pääsenkö minä kanssasi?" kysyi Anna Maija, tullen juuri


keskeyttämään toisten tilisalaisuutta.

"Mitäpä sinulla on kaupungissa asiaa? Suotta vain lähtisit hevosen


vedettäväksi."

Seuraavana aamuna aikasin Anna Maija kuitenkin oli vaatteissaan


ja istahti Matin rekeen, johon emäntä sill'aikaa, kuin Matti vielä
ennen lähtöään pistihe tupaan piippuansa sytyttämään, toi salaa
muilta paitsi Anna Maijalta isonlaisen sijan reisikappaleen
kaupungissa tavaraan vaihdettavaksi.

Illan suussa palasi kaupunkimatkue.

Anna Maija juoksi hyvin tyytymättömänä ennen muita tupaan,


josta äitinsä heti vastaan tullen yritti käännyttämään häntä takaisin
porstuaan. Mutta Anna Maija vain äreästi alkoi riisuutua.

Korjattuaan hevosen saapui vähän ajan perästä Mattikin, kantaen


sijankinkkua kintusta, ja Mikko ilvehtien riensi jälestä.

"Katsokaas, millaisen lihakimpaleen minä siellä kaupungissa


löysin meidän reestä! Eikä sitä kukaan omistanut, vaikka minä kyllä
ilmoittelin."

"Eiköhän vain liene omia lihavaroja. Hyvinhän näyttikin


lihakorvossa suuri kolo olevan, kuin kävin siitä ottamassa
paistipalasta."

Liena se näin suoraan lausui arvelujansa, luullen vain Anna Maijaa


yksin syylliseksi, hän kun oli puoli väkisin sulloutunut lähtemään
kaupunkiin.

"Mikäs sen sinne kaupunkiin toi?" ilvensi Mikko.

"Eikö tuo lie hevonen vetänyt", tarttui nyt isäntä puheesen ja katsoi
terävästi Anna Maijaan, luullen hänkin tytärtänsä yksin syypääksi.

"No, hevonen sen tietysti vei, kun se täältä pantiin menemään.


Mutta miks'et sitä myönyt, Anna Maija?"

Tällä selityksellään emäntä tahtoi pelastaa tytärtänsä pulasta.

Anna Maija ryntäsi itku kurkussa ulos.

"Eikö sinulla ole muita kaupungin tuomisia?" kysyi isäntä Matilta,


ikään kuin ei olisi huomannutkaan emännän ilmoitusta. "Vai
ihastuitko sinä niin lihan löydöstä, että unhotit kaikki ostokset
sikseen?"
Matti pistäysi noutamaan ison kontin ja kaivoi siitä esiin pussin
toisensa perästä.

"En toki jättänyt; on niitä kontissa. Tässä on herrasryyniä, tässä


vehnäjauhoja, tässä kahvia ja tässä koko pikku pää sokuria. Onpa
tuossa teetäkin. Ja neuloja koko ropakko, kun kuulin äidin niitä
kaipaavan. No, Jussi, tulepas tänne osalle. Kas, tästä saat koko
kahmalosi täyteen oikeita sokurikakkuja. Mutta menepäs nyt
tarjoamaan isälle ja äidille; Lienan ja Anna Maijan pitää myöskin
saada, ja tehneehän se Mikonkin mieli, vaikka hänelle olisi parahiksi,
jos jäisikin ilman, kun jo kävi omin lupinsa pussissa."

"No, Anni, eikö näistä sinulle riitä?" kysyi isäntä.

"Kukapa tuota arvasi teidän niitä ostelevan."

"Senkö tähden piti niitä salaa muka hommata? Ikään kuin minä en
vanhastaan tietäisi, että vehnäsiä kuitenkin tavalla tai toisella laitat
jouluksi. Heitä sinä pois se salaamisen puuha. Nyt meillä on
kuitenkin monet silmät näkemässä.”

"Näkyypä olevan ja varsinkin Lienalla”, tuli emäntä äkeissään


lausuneeksi, ehtimättä ajatella, että ne sanat olisivat saattaneet
pahentaa hänen ja miniänsä väliä.

"Älä syytä Lienaa, kun on syy itsessäsi. Sano, että parempaa


miniää et olisi mistään saanut. Vai eikö pysty töihin taikka tekeekö
mitään salaa?"

Emännän täytyi mielessään väkisinkin tunnustaa isännän sana


todeksi, mutta julki hän ei sitä lausunut, eikä isäntäkään häntä enää

You might also like