Ebook
Applied Univariate, Bivariate, and Multivariate StatisticsISBN: 9781118632239
760 pages
October 2015

Description
A clear and efficient balance between theory and application of statistical modeling techniques in the social and behavioral sciences
Written as a general and accessible introduction, Applied Univariate, Bivariate, and Multivariate Statistics provides an overview of statistical modeling techniques used in fields in the social and behavioral sciences. Blending statistical theory and methodology, the book surveys both the technical and theoretical aspects of good data analysis.
Featuring applied resources at various levels, the book includes statistical techniques such as ttests and correlation as well as more advanced procedures such as MANOVA, factor analysis, and structural equation modeling. To promote a more indepth interpretation of statistical techniques across the sciences, the book surveys some of the technical arguments underlying formulas and equations. Applied Univariate, Bivariate, and Multivariate Statistics also features
 Demonstrations of statistical techniques using software packages such as R and SPSS^{}
 Examples of hypothetical and real data with subsequent statistical analyses
 Historical and philosophical insights into many of the techniques used in modern social science
 A companion website that includes further instructional details, additional data sets, solutions to selected exercises, and multiple programming options
An ideal textbook for courses in statistics and methodology at the upper undergraduate and graduatelevels in psychology, political science, biology, sociology, education, economics, communications, law, and survey research, Applied Univariate, Bivariate, and Multivariate Statistics is also a useful reference for practitioners and researchers in their field of application.
DANIEL J. DENIS, PhD, is Associate Professor of Quantitative Psychology at the University of Montana where he teaches courses in univariate and multivariate statistics. He has published a number of articles in peerreviewed journals and has served as consultant to researchers and practitioners in a variety of fields.
Table of Contents
Preface xix
About the Companion Website xxxiii
1 Preliminary Considerations 1
1.1 The Philosophical Bases of Knowledge: Rationalistic versus Empiricist Pursuits 1
1.2 What is a “Model”? 4
1.3 Social Sciences versus Hard Sciences 6
1.4 Is Complexity a Good Depiction of Reality? Are Multivariate Methods Useful? 8
1.5 Causality 9
1.6 The Nature of Mathematics: Mathematics as a Representation of Concepts 10
1.7 As a Social Scientist How Much Mathematics Do You Need to Know? 11
1.8 Statistics and Relativity 12
1.9 Experimental versus Statistical Control 13
1.10 Statistical versus Physical Effects 14
1.11 Understanding What “Applied Statistics” Means 15
Review Exercises 15
2 Mathematics and Probability Theory 18
2.1 Set Theory 20
2.2 Cartesian Product A × B 24
2.3 Sets of Numbers 26
2.4 Set Theory Into Practice: Samples Populations and Probability 27
2.5 Probability 28
2.6 Interpretations of Probability: Frequentist versus Subjective 35
2.7 Bayes’ Theorem: Inverting Conditional Probabilities 39
2.8 Statistical Inference 44
2.9 Essential Mathematics: Precalculus Calculus and Algebra 48
2.10 Chapter Summary and Highlights 72
Review Exercises 74
3 Introductory Statistics 78
3.1 Densities and Distributions 79
3.2 ChiSquare Distributions and GoodnessofFit Test 91
3.3 Sensitivity and Specificity 98
3.4 Scales of Measurement: Nominal Ordinal and Interval Ratio 98
3.5 Mathematical Variables versus Random Variables 101
3.6 Moments and Expectations 103
3.7 Estimation and Estimators 106
3.8 Variance 108
3.9 Degrees of Freedom 110
3.10 Skewness and Kurtosis 111
3.11 Sampling Distributions 113
3.12 Central Limit Theorem 116
3.13 Confidence Intervals 117
3.14 Bootstrap and Resampling Techniques 119
3.15 Likelihood Ratio Tests and Penalized LogLikelihood Statistics 121
3.16 Akaike’s Information Criteria 122
3.17 Covariance and Correlation 123
3.18 Other Correlation Coefficients 128
3.19 Student’s t Distribution 131
3.20 Statistical Power 139
3.21 Paired Samples tTest: Statistical Test for Matched Pairs (Elementary Blocking) Designs 146
3.22 Blocking with Several Conditions 149
3.23 Composite Variables: Linear Combinations 149
3.24 Models in Matrix Form 151
3.25 Graphical Approaches 152
3.26 What Makes a pValue Small? A Critical Overview and Simple Demonstration of Null Hypothesis
Significance Testing 155
3.27 Chapter Summary and Highlights 164
Review Exercises 167
4 Analysis of Variance: Fixed Effects Models 173
4.1 What is Analysis of Variance? Fixed versus Random Effects 174
4.2 How Analysis of Variance Works: A Big Picture Overview 178
4.3 Logic and Theory of ANOVA: A Deeper Look 180
4.4 From Sums of Squares to Unbiased Variance Estimators: Dividing by Degrees of Freedom 189
4.5 Expected Mean Squares for OneWay Fixed Effects Model: Deriving the FRatio 190
4.6 The Null Hypothesis in ANOVA 196
4.7 Fixed Effects ANOVA: Model Assumptions 198
4.8 A Word on Experimental Design and Randomization 201
4.9 A Preview of the Concept of Nesting 201
4.10 Balanced versus Unbalanced Data in ANOVA Models 202
4.11 Measures of Association and Effect Size in ANOVA: Measures of Variance Explained 202
4.12 The FTest and the Independent Samples tTest 205
4.13 Contrasts and PostHocs 205
4.14 PostHoc Tests 212
4.15 Sample Size and Power for ANOVA: Estimation with R and G∗Power 218
4.16 Fixed Effects OneWay Analysis of Variance in R: Mathematics Achievement as a Function of Teacher 222
4.17 Analysis of Variance Via R’s lm 226
4.18 Kruskal–Wallis Test in R 227
4.19 ANOVA in SPSS: Achievement as a Function of Teacher 228
4.20 Chapter Summary and Highlights 230
Review Exercises 232
5 Factorial Analysis of Variance: Modeling Interactions 237
5.1 What is Factorial Analysis of Variance? 238
5.2 Theory of Factorial ANOVA: A Deeper Look 239
5.3 Comparing OneWay ANOVA to TwoWay ANOVA: Cell Effects in Factorial ANOVA versus Sample
Effects in OneWay ANOVA 245
5.4 Partitioning the Sums of Squares for Factorial ANOVA: The Case of Two Factors 246
5.5 Interpreting Main Effects in the Presence of Interactions 253
5.6 Effect Size Measures 253
5.7 ThreeWay FourWay and HigherOrder Models 254
5.8 Simple Main Effects 254
5.9 Nested Designs 256
5.9.1 Varieties of Nesting: Nesting of Levels versus Subjects 257
5.10 Achievement as a Function of Teacher and Textbook: Example of Factorial ANOVA in R 258
5.11 Interaction Contrasts 266
5.12 Chapter Summary and Highlights 267
Review Exercises 268
6 Introduction to Random Effects and Mixed Models 270
6.1 What is Random Effects Analysis of Variance? 271
6.2 Theory of Random Effects Models 272
6.3 Estimation in Random Effects Models 273
6.4 Defining Null Hypotheses in Random Effects Models 276
6.5 Comparing Null Hypotheses in Fixed versus Random Effects Models: The Importance of Assumptions
278
6.6 Estimating Variance Components in Random Effects Models: ANOVA ML REML Estimators 279
6.7 Is Achievement a Function of Teacher? OneWay Random Effects Model in R 282
6.8 R Analysis Using REML 285
6.9 Analysis in SPSS: Obtaining Variance Components 286
6.10 Factorial Random Effects: A TwoWay Model 287
6.11 Fixed Effects versus Random Effects: A Way of Conceptualizing Their Differences 289
6.12 Conceptualizing the TwoWay Random Effects Model: The Makeup of a Randomly Chosen Observation 289
6.13 Sums of Squares and Expected Mean Squares for Random Effects: The Contaminating Influence of Interaction Effects 291
6.14 You Get What You Go in with: The Importance of Model Assumptions and Model Selection 293
6.15 Mixed Model Analysis of Variance: Incorporating Fixed and Random Effects 294
6.16 Mixed Models in Matrices 298
6.17 Multilevel Modeling as a Special Case of the Mixed Model: Incorporating Nesting and Clustering 299
6.18 Chapter Summary and Highlights 300
Review Exercises 301
7 Randomized Blocks and Repeated Measures 303
7.1 What Is a Randomized Block Design? 304
7.2 Randomized Block Designs: Subjects Nested Within Blocks 304
7.3 Theory of Randomized Block Designs 306
7.4 Tukey Test for Nonadditivity 311
7.5 Assumptions for the Variance–Covariance Matrix 311
7.6 Intraclass Correlation 313
7.7 Repeated Measures Models: A Special Case of Randomized Block Designs 314
7.8 Independent versus Paired Samples tTest 315
7.9 The Subject Factor: Fixed or Random Effect? 316
7.10 Model for OneWay Repeated Measures Design 317
7.11 Analysis Using R: OneWay Repeated Measures: Learning as a Function of Trial 318
7.12 Analysis Using SPSS: OneWay Repeated Measures: Learning as a Function of Trial 322
7.13 SPSS: TwoWay Repeated Measures Analysis of Variance: Mixed Design: One Between Factor One
Within Factor 326
7.14 Chapter Summary and Highlights 330
Review Exercises 331
8 Linear Regression 333
8.1 Brief History of Regression 334
8.2 Regression Analysis and Science: Experimental versus Correlational Distinctions 336
8.3 A Motivating Example: Can Offspring Height Be Predicted? 337
8.4 Theory of Regression Analysis: A Deeper Look 339
8.5 Multilevel Yearnings 342
8.6 The LeastSquares Line 342
8.7 Making Predictions Without Regression 343
8.8 More About εi 345
8.9 Model Assumptions for Linear Regression 346
8.10 Estimation of Model Parameters in Regression 349
8.11 Null Hypotheses for Regression 351
8.12 Significance Tests and Confidence Intervals for Model Parameters 353
8.13 Other Formulations of the Regression Model 355
8.14 The Regression Model in Matrices: Allowing for More Complex Multivariable Models 356
8.15 Ordinary LeastSquares in Matrices 359
8.16 Analysis of Variance for Regression 360
8.17 Measures of Model Fit for Regression: How Well Does the Linear Equation Fit? 363
8.18 Adjusted R2 364
8.19 What “Explained Variance” Means: And More Importantly What It Does Not Mean 364
8.20 Values Fit by Regression 365
8.21 LeastSquares Regression in R: Using Matrix Operations 365
8.22 Linear Regression Using R 368
8.23 Regression Diagnostics: A Check on Model Assumptions 370
8.24 Regression in SPSS: Predicting Quantitative from Verbal 379
8.25 Power Analysis for Linear Regression in R 383
8.26 Chapter Summary and Highlights 384
Review Exercises 385
9 Multiple Linear Regression 389
9.1 Theory of Partial Correlation and Multiple Regression 390
9.2 Semipartial Correlations 392
9.3 Multiple Regression 393
9.4 Some Perspective on Regression Coefficients: “Experimental Coefficients”? 394
9.5 Multiple Regression Model in Matrices 395
9.6 Estimation of Parameters 396
9.7 Conceptualizing Multiple R 396
9.8 Interpreting Regression Coefficients: The Case of Uncorrelated Predictors 397
9.9 Anderson’s IRIS Data: Predicting Sepal Length from Petal Length and Petal Width 397
9.10 Fitting Other Functional Forms: A Brief Look at Polynomial Regression 402
9.11 Measures of Collinearity in Regression: Variance Inflation Factor and Tolerance 403
9.12 RSquared as a Function of Partial and Semipartial Correlations: The Stepping Stones to Forward and Stepwise Regression 405
9.13 ModelBuilding Strategies: Simultaneous Hierarchichal Forward and Stepwise 406
9.14 Power Analysis for Multiple Regression 410
9.15 Introduction to Statistical Mediation: Concepts and Controversy 411
9.16 Chapter Summary and Highlights 414
Review Exercises 415
10 Interactions in Multiple Linear Regression: Dichotomous Polytomous and Continuous Moderators 418
10.1 The Additive Regression Model with Two Predictors 420
10.2 Why the Interaction is the Product Term xizi: Drawing an Analogy to Factorial ANOVA 420
10.3 A Motivating Example of Interaction in Regression: Crossing a Continuous Predictor with a
Dichotomous Predictor 421
10.4 Theory of Interactions in Regression 424
10.5 Simple Slopes for Continuous Moderators 427
10.6 A Simple Numerical Example: How Slopes Can Change as a Function of the Moderator 428
10.7 Calculating Simple Slopes: A Useful Algebraic Derivation 430
10.8 Summing Up the Idea of Interactions in Regression 432
10.9 Do Moderators Really “Moderate” Anything? Some Philosophical Considerations 432
10.10 Interpreting Model Coefficients in the Context of Moderators 433
10.11 MeanCentering Predictors: Improving the Interpretability of Simple Slopes 434
10.12 The Issue of Multicollinearity: A Second Reason to Like MeanCentering 435
10.13 Interaction of Continuous and Polytomous Predictors in R 436
10.14 Multilevel Regression: Another Special Case of the Mixed Model 440
10.15 Chapter Summary and Highlights 441
Review Exercises 441
11 Logistic Regression and the Generalized Linear Model 443
11.1 Nonlinear Models 445
11.2 Generalized Linear Models 447
11.3 Canonical Links 450
11.4 Distributions and Generalized Linear Models 451
11.5 Dispersion Parameters and Deviance 453
11.6 Logistic Regression: A Generalized Linear Model for Binary Responses 454
11.7 Exponential and Logarithmic Functions 456
11.8 Odds Odds Ratio and the Logit 461
11.9 Putting It All Together: The Logistic Regression Model 462
11.10 Logistic Regression in R: Challenger ORing Data 466
11.11 Challenger Analysis in SPSS 469
11.12 Sample Size Effect Size and Power 473
11.13 Further Directions 474
11.14 Chapter Summary and Highlights 475
Review Exercises 476
12 Multivariate Analysis of Variance 479
12.1 A Motivating Example: Quantitative and Verbal Ability as a Variate 480
12.2 Constructing the Composite 482
12.3 Theory of MANOVA 482
12.4 Is the Linear Combination Meaningful? 483
12.5 Multivariate Hypotheses 487
12.6 Assumptions of MANOVA 488
12.7 Hotelling’s T2: The Case of Generalizing from Univariate to Multivariate 489
12.8 The Variance–Covariance Matrix S 492
12.9 From Sums of Squares and CrossProducts to Variances and Covariances 494
12.10 Hypothesis and Error Matrices of MANOVA 495
12.11 Multivariate Test Statistics 495
12.12 Equality of Variance–Covariance Matrices 500
12.13 Multivariate Contrasts 501
12.14 MANOVA in R and SPSS 502
12.15 MANOVA of Fisher’s Iris Data 508
12.16 Power Analysis and Sample Size for MANOVA 509
12.17 Multivariate Analysis of Covariance and Multivariate Models: A Bird’s Eye View of Linear Models 511
12.18 Chapter Summary and Highlights 512
Review Exercises 513
13 Discriminant Analysis 517
13.1 What is Discriminant Analysis? The Big Picture on the Iris Data 518
13.2 Theory of Discriminant Analysis 520
13.3 LDA in R and SPSS 523
13.4 Discriminant Analysis for Several Populations 529
13.5 Discriminating Species of Iris: Discriminant Analyses for Three Populations 532
13.6 A Note on Classification and Error Rates 535
13.7 Discriminant Analysis and Beyond 537
13.8 Canonical Correlation 538
13.9 Motivating Example for Canonical Correlation: Hotelling’s 1936 Data 539
13.10 Canonical Correlation as a General Linear Model 540
13.11 Theory of Canonical Correlation 541
13.12 Canonical Correlation of Hotelling’s Data 544
13.13 Canonical Correlation on the Iris Data: Extracting Canonical Correlation from Regression MANOVA
LDA 546
13.14 Chapter Summary and Highlights 547
Review Exercises 548
14 Principal Components Analysis 551
14.1 History of Principal Components Analysis 552
14.2 Hotelling 1933 555
14.3 Theory of Principal Components Analysis 556
14.4 Eigenvalues as Variance 557
14.5 Principal Components as Linear Combinations 558
14.6 Extracting the First Component 558
14.7 Extracting the Second Component 560
14.8 Extracting Third and Remaining Components 561
14.9 The Eigenvalue as the Variance of a Linear Combination Relative to Its Length 561
14.10 Demonstrating Principal Components Analysis: Pearson’s 1901 Illustration 562
14.11 Scree Plots 566
14.12 Principal Components versus LeastSquares Regression Lines 569
14.13 Covariance versus Correlation Matrices: Principal Components and Scaling 570
14.14 Principal Components Analysis Using SPSS 570
14.15 Chapter Summary and Highlights 575
Review Exercises 576
15 Factor Analysis 579
15.1 History of Factor Analysis 580
15.2 Factor Analysis: At a Glance 580
15.3 Exploratory versus Confirmatory Factor Analysis 581
15.4 Theory of Factor Analysis: The Exploratory FactorAnalytic Model 582
15.5 The Common FactorAnalytic Model 583
15.6 Assumptions of the FactorAnalytic Model 585
15.7 Why Model Assumptions Are Important 587
15.8 The Factor Model as an Implication for the Covariance Matrix Σ 587
15.9 Again Why is Σ ˆ ΛΛ´ ‡ ψ so Important a Result? 589
15.10 The Major Critique Against Factor Analysis: Indeterminacy and the Nonuniqueness of Solutions 589
15.11 Has Your Factor Analysis Been Successful? 591
15.12 Estimation of Parameters in Exploratory Factor Analysis 592
15.13 Estimation of Factor Scores 593
15.14 Principal Factor 593
15.15 Maximum Likelihood 595
15.16 The Concepts (and Criticisms) of Factor Rotation 596
15.17 Varimax and Quartimax Rotation 599
15.18 Should Factors Be Rotated? Is That Not “Cheating?” 600
15.19 Sample Size for Factor Analysis 601
15.20 Principal Components Analysis versus Factor Analysis: Two Key Differences 602
15.21 Principal Factor in SPSS: Principal Axis Factoring 604
15.22 Bartlett Test of Sphericity and Kaiser–Meyer–Olkin Measure of Sampling Adequacy (MSA) 612
15.23 Factor Analysis in R: Holzinger and Swineford (1939) 613
15.24 Cluster Analysis 616
15.25 What Is Cluster Analysis? The Big Picture 617
15.26 Measuring Proximity 619
15.27 Hierarchical Clustering Approaches 623
15.28 Nonhierarchical Clustering Approaches 625
15.29 KMeans Cluster Analysis in R 626
15.30 Guidelines and Warnings About Cluster Analysis 630
15.31 Chapter Summary and Highlights 630
Review Exercises 632
16 Path Analysis and Structural Equation Modeling 636
16.1 Path Analysis: A Motivating Example—Predicting IQ Across Generations 637
16.2 Path Analysis and “Causal Modeling” 639
16.3 Early PostWright Path Analysis: Predicting Child’s IQ (Burks 1928) 641
16.4 Decomposing Path Coefficients 642
16.5 Path Coefficients and Wright’s Contribution 644
16.6 Path Analysis in R: A Quick Overview—Modeling Galton’s Data 644
16.7 Confirmatory Factor Analysis: The Measurement Model 648
16.8 Structural Equation Models 650
16.9 Direct Indirect and Total Effects 652
16.10 Theory of Statistical Modeling: A Deeper Look into Covariance Structures and General Modeling 653
16.11 Other Discrepancy Functions 655
16.12 The Discrepancy Function and ChiSquare 656
16.13 Identification 657
16.14 Disturbance Variables 659
16.15 Measures and Indicators of Model Fit 660
16.16 Overall Measures of Model Fit 660
16.17 Model Comparison Measures: Incremental Fit Indices 662
16.18 Which Indicator of Model Fit Is Best? 665
16.19 Structural Equation Model in R 666
16.20 How All Variables Are Latent: A Suggestion for Resolving the Manifest–Latent Distinction 668
16.21 The Structural Equation Model as a General Model: Some Concluding Thoughts on Statistics and
Science 669
16.22 Chapter Summary and Highlights 670
Review Exercises 671
Appendix A: Matrix Algebra 675
References 705
Index 721