Resources
Statistical Programs
Statistical Primers
Statistical Primers
Interfaces with SPSS, Excel, Ascii, SAS and STATA Formats
This page contains resources for each book chapter, including links of potential interest, additional documents, and data sets used in the chapters.
I make available here a power point file with every figure from the book on a separate slide. You can select and then copy and paste figures of your choosing into your presentations, lectures or handouts.
CHAPTER 1: RANDOMIZED EXPLANATORY TRIALS
1. For answers to frequently asked questions for this chapter topic, click here.
CHAPTER 2: CONCEPTUAL FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For drawing influence diagrams, a flexible program is Microsoft Visio, which is available through many universities. It also is available at a reasonable cost through third party vendors, like Brytesoft. Click below for a tutorial on how to use this program and a description of tools for use in Visio.
Read a short tutorial on how to use Visio.
Download templates and vss files discussed in the tutorial (this is a zip file that, when opened, will make 4 files available). You may have to adjust your Trust Settings in Visio to use these files - see the tutorial.
Another viable graphics program for influence diagrams that is free is app.diagrams. Click here for it.
3. For use on your computer without an internet connection, you can download the program that the daggity.net website uses to identify confounds, independence relations, and instrumental variables here. A pdf manual for the program is here.
a. A free, self-paced course on using DAGs in causal analysis is offered by Harvard University here.
b. DAGs conceptualize latent variables somewhat differently than SEM; latent variables are unobserved variables with no indicators whereas in SEM, latent variables often (but do not always) have reference indicators. For a technical discussion of the role of latent variables in DAGs, click here.
c. A good introductory treatment of DAGs is in Pearl, Glymour, and Jewell (2016).
4. For a copy of the Cinelli, Forney and Pearl (2022) paper on good and bad controls and arguments against atheoretical partialling, click here.
5. For a document on collider bias and mediation analysis, click here
CHAPTER 3: MEASUREMENT FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a brief overview of key concepts related to scaling theory for the construction of multi-item scales, click here.
3. For a primer on testing for unidimensionality of a multi-item scale, click here.
4. For a primer on strategies for addressing measurement error for single indicator SEM models, click here.
5. For a primer on testing measurement invariance, click here. To download the Mplus data file used in the primer, click here. Use the following names for the variables in this order in the Mplus syntax: id d1 d2 d3 d4 d5 d6 adhere adhere2 dfemale income ethnic.
6. For an open access article on a brief social desirability scale (that includes the scale itself), click here. I sometimes adapt the scale based on my study context and population.
7. For the chapter from Jaccard and Jacoby (2020) that describes application of the framework focused on the processes of comprehension, judgment, and response translation to measure construction, click here.
CHAPTER 4: METHODOLOGICAL FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For an Amazon link to used books of the classic book on experimental and quasi-experimental design by Campbell and Stanley, click here. This is well worth a read.
3. For an updated (but still somewhat old) version of the Campbell and Stanley book (a lengthy chapter by Sadish, Campbell and Stanley), click here. It is downloadable from the internet.
4. For a more recent and excellent text on experimental design and quasi experimental design by Charles Reichardt titled "Quasi-Experimentation: A Guide to Design and Analysis", click here. There is a wealth of information about experimental design per se.
5. For a youtube introductory video on common clincial trial designs, click here.
CHAPTER 5: STATISTICAL FUNDAMENTALS FOR RETs: REGRESION
1. For answers to frequently asked questions for this chapter topic, click here.
2. There are a large number of books on the different regression methods discussed in Chapter 5 making recommendations challenging. Sage publishes a series of relatively introductory, inexpensive monographs, referred to by some as the "little green books" that cover many statistical topics, including regression. For a list of the regression books in the Sage series, click here. The SAGE regression webpage includes links in the "search by category" box for other little green books.
3. A readable competing series of introductory monographs (called "Blue Books") is offered by David Garson. Many of these monographs are available on Amazon as ebooks and they are free. For a list of these monographs, click here. Be sure to focus on from this list those monographs that are in the Blue Book series.
4. Guilford Publications offers high quality, readable and reasonably priced paperbacks on a range of regression related topics as part of their series "Methodology and Statistics". For a link to the books in this series, click here. Lot's of great books in this series, so it is worth checking out.
CHAPTER 6: STATISTICAL FUNDAMENTALS FOR RETs: ADVANCED TOPICS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a primer on smoothers, click here.
3. For a primer on quantile regression, click here.
4. For an introductory youtube video on instrumental variables by the Nobel Prize Laureate Joshua Angrist, click here. This is one video in a series of videos on econometrics.
5. For an introductory youtube video on margins of error, click here.
6. For an introductory youtube video on centering, click here.
7. Rand Wilcox at USC is an expert on robust analyses and has written numerous books on them that range from clearly written introductory treatments to more advanced treatments. I highly recommend his books and articles. For his webpage, click here. Be sure to click on the link to his CV on this webpage.
8. For an introduction to bootstrapping and two useful introductory videos to bootstrapping, click here.
CHAPTER 7: STATISTICAL FUNDAMENTALS FOR RETs: STRUCTURAL EQUATION MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a video of the SEM decomposition process described in Chapter 7, click here.
3. For a useful website by Dave Kenny on many SEM topics, click here.
4. To access a popular SEM blog called SEMNET, click here. Two other potentially useful blogs for SEM advice include the blog for the lavaan Google group (click here) and a blog called CrossValidated that includes more general statistical questions and answers (click here).
5. For the webpage of Rex Kline's introductory book on SEM, click here.
6. For the webpage of Tim Brown's book on confirmatory factor analysis, click here.
7. For the webpage of Rick Hoyle's Handbook of SEM, click here.
8. For a document that describes SEM estimation algorithms in Mplus, click here.
9. For a youtube video on fit indices in SEM, click here.
10. Patrick Curran of the University of North Carlolina provides a free 6 you tube video series on SEM as part of CenterStat. To access these videos (SEM Episodes 1 to 6), click here. There are multiple other videos on this web page you may find of interest, such as a nine video treatment of growth curve modeling and some videos (by Dan Bauer) on latent class mixture models
and a seven video series on multiple regression.
11. Mikko Rönkkö has published on youtube over 200 videos on different facets of SEM. To go to his youtube page for the videos, click here.
CHAPTER 8: STATISTICAL FUNDAMENTALS FOR RETs: NON-TRADITIONAL SEM
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a pdf of the classic book on Bayesian Data Analysis by Gelman et al., click here.
3. For a website on piecewise SEM and an R package for it, click here.
4. For a document that describes how to apply quantile regression to RETs using limited information SEM, click here.
5. For a youtube video about Bollen's model implied instrumental variable approach, click here.
7. For a useful powerpoint on the use of Bollen's model implied instrumental variable approach, click here.
8. For Judea Pearl's homepage with many resources for Structural Causal Modeling, click here.
9. For a primer on how to locate the asymptotic covariance matrix in Mplus, lavaan, MIIVsem, STATA, R, and SPSS output for the calculation of Monte Carlo confidence intervals, click here.
CHAPTER 9: MEDIATION ANALYSIS IN RETs: BASIC APPROACHES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a consensus panel's recommendations on what to report for mediation analyses, click here.
3. For the web page on mediation on the Mplus website, click here.
4. For Tyler VanderWeele's homepage with useful videos on casual mediation frameworks, click here.
CHAPTER 10: EVALUATING EFFECT SIZES IN RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a web site to calculate traditional standardized effect sizes, click here. The probability of exceptions to the rule index, expressed as a percentage, can be calculated by subtracting the CLES (common language effect size) index on this website from 100. To convert one effect size index to another effect size index, see entry number 14 called Transformation of the effect sizes d, r, f, odds ratio, eta squared, and CLES. The site also has a nice section on testing the significance of correlations that implements many tests of differences between correlations for independent and dependent groups.
3. For an interactive, graphical illustration of the relationship of Cohen's d to overlap indices and the probability of exceptions to the rule, click here. This website calls the common language effect size the probability of superiority and expresses it in percentage form. Subtracting it from 100 yields the probability of exceptions to the rule, expressed as a percentage. The website includes an index of the number needed to treat (NNT) for continuous outcomes using a cutoff score to dichotomize the continuous outcome, which, in my opinion, is dubious because it falsely dichotomizes a continuous construct. The website allows you to alter the default value of the cutoff (see the formulas section of the website). Having said that, if you work with a substantively defined, theoretically meaningful cutoff value, this is a reasonable tool to graphically explore the relationship between d and NNT with cutoffs.
4. For a primer on how to conduct hierarchical regression analysis in Mplus in order to obtain indices of unique explained variance using robust estimators such as MLR and how to use MM regression to isolate outlier resistant indices of incremental explained variance, click here.
CHAPTER 11: MEDIATION ANALYSIS WITH CONTINUOUS OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a video that talks you through Mplus syntax for the numerical example in Chapter 11, click here.
3. For a document that describes preliminary analyses for the social phobia example, click here.
4. A misunderstood warning message that sometimes occurs in Mplus reads "THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION." The warning appears in the output of the example for Chapter 11. Sometimes this warning requires model modification and sometimes it can reasonably be ignored. For an Mplus document that describes the technical bases of the message, click here. The warning can be safely ignored in the example for Chapter 11. In my opinion, the Mplus document sometimes creates the impression that corrective action might be needed in certain cases where it probably is not, but the presence of the message in and of itself should cause you to reflect on the matter and ensure you are comfortable moving forward, perhaps after looking at further diagnostics.
5. For a document on measurement non-invariance tests for the latent social phobia indicators in the chapter example, click here.
6. For a document that describes how to calculate standardized effect sizes for the chapter example, click here.
7. For a document that shows how to use Mplus to perform effect decompositions, click here.
8. For a document that shows how to calculate or obtain from Mplus classic omnibus mediation tests and interventional indirect effects for the social phobia example, click here.
9. For a document that describes how to conduct sensitivity tests in full information SEM, click here.
10. For a document that describes how to deal with unmeasured confounds and correlated distrubances using the social phobia example, click here.
11. For a document that shows how to get basic descriptive statistics in Mplus using the BASIC option, click here.
12. For a document that describes tests for competing models to the social phobia model, click here.
13. For a document that shows how to adjust single indicators for measurement error in the chapter example, click here.
14. For an annotated output of the chapter example using traditional FISEM, click here.
15. For the DAGitty code to identify independence tests for the social phobia example, click here.
16. For R-compatible (chap11R.txt) and Mplus-compatible (chap11M.txt) data files for the social phobia example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id cr1 spai1 spin1 cr3 spai3 spin3 negapp2 pskills2 extern2 negapp1 pskills1 extern1 hyper sex treat
17. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 12: MEDIATION ANALYSIS WITH BINARY OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a document that describes preliminary analyses for the parent communication example, click here
3. For the DAGitty code to identify independence tests for the parent communication example, click here.
4. . For a document that shows how to calculate or obtain from Mplus the classic omnibus mediation tests for the parent communication example, click here. This includes material on omnibus tests for the causal mediation framework.
5. For a document that shows how to calculate margins of error and confidence intervals for indices of relative risk and the number needed to treat using Mplus, click here.
6. For programming the chapter example using logit instead of probit modeling, click here.
7. For annotated output of the chapter example using traditional probit-based modeling, click here.
8. For an informative video on the shortcomings of logistic regression by Edward Norton, a leading economist, click here. This video is very slow to download so be patient. Also check out the handouts and transcript for the video on that same webpage.
9. For the classic article by Mood critiquing logistic regression, click here. For an article by Gomila comparing logistic regression and the linear probability model, click here and for a critque of logistic regression by Norton, click here.
10. For R-compatible (chap12R.txt) and Mplus-compatible (chap12M.txt) data files for the parent communication example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id com3 pa2 pk2 pe2 cq1 pa1 pk1 pe1 t bs1
11. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 13: MEDIATION ANALYSIS WITH ORDINAL AND NOMINAL OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on ordinal regression, click here.
3. For a useful youtube video on how to compute marginal effects for predictors in ordinal regression using STATA, click here.
4. For a youtube video on working with ordinal level predictors, click here. In the video, I do not agree with the assertion that all of the coefficients for staircase based dummy variables need to be statistically signficant. I believe that as long as at least one is statistically signficant that this suggests something is going on. But aside from this quibble, the video is pretty good.
5. For annotated output of the chapter example using traditional probit-based analysis, click here.
6. For R-compatible (chap13R.txt) and Mplus-compatible (chap13M.txt) data files for the clinician training example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): ID GA2 TA2 BD2 CE1 CIS1 TREAT IMP3 y234 y34 y4 cCE1 cCIS1.
7. For a document called Supplenetary Total Efffects Analysis for Ordinal and Multinomial Modeling that provides a nonparametric method for analyzing total effects in ordinal or multinomial regression, click here.
8. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 14: MEDIATION ANALYSIS WITH COUNT OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
CHAPTER 15: NON-LINEAR AND SPECIALIZED MODELING IN MEDIATION ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on quadratic regression, click here.
3. For a youtube video on regression splines, click here.
4. For an introductory book chapter by Jaccard on how to build math models using traditional non-linear regression approaches (and more), click here.
5. For two useful youtube videos on Bayesian Additive Regression Trees (BART), click here and here. These videos develop the mathematical details of BART in more detail than I do in my chapter but are still somewhat conceptual in focus.
6. For introductory youtube videos on generalized additive models by Noam Ross, click here, here and here. For an additional workshop video by Noam, click here. For a free inteactive course on GAMs by Noam, click here.
7. For a brief primer on cluster plots when a cluster analysis focuses on more than two target variables, click here.
8. For an introductory youtube video on principal component analysis, click here
9. For an introductory youtube video on discriminant analysis, click here
10. The R program I use for variants of trimmed k-means analysis is tclust. It is dated. A newer version of tclust that offers more flexibility and features requires special downloading steps than is typical in R. For instructions, click here. The program I use on my website is the older version. You only need to download the newer version if you are going to do more advanced R programming on your own independent of my program. Otherwise you can just wait until the newer version makes it to CRAN.
11. For R-compatible (quadraticR.txt) and Mplus-compatible (quadraticM.txt) data files for the quadratic RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): treat m1 m2 m3 m1m1 y v1 v2 v3 v4. Exclude variables v1-v4 with the USEVARIABLES command within Mplus.
12. For R-compatible (splineR.txt) and Mplus-compatible (splineM.txt) data files for the spline modeling RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id m1 m2 y treat covm1 covm2 covy ccovy.
13. For R-compatible (nonlinearR.txt) and Mplus-compatible (nonlinearR.txt) data files for the traditional non-linear regression RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id y m1 m2 treat.
14. For R-compatible (bartdat.txt) data file for the BART example, click here.
15. For R-compatible (rcluster2R.txt) and Mplus-compatible (rcluster2M.txt) data files for the trimmed k-means robust cluster analysis RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id warmth control clus treat d1clus d2clus d3clus d4clus cov1 cov2 y. Exclude variables from the analysis with the USEVARIABLES command, as appropriate.
16. For a document that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 16: RETs WITH FOLLOW-UPS: LONGITUDINAL MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a useful article on lagged dependent variables, click here. For a useful discussion by Paul Allison on the use of lagged dependent variables in mixed models, click here.
3. For a document that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 17: MEDIATOR RELATIVE IMPORTANCE AND EXPLORATORY MEDIATOR ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a link to the relaimpo package in R for relative importance analysis, click here.
3. For a link to conducting dominance analysis using STATA, click here.
4. For a worked example of mediator selection using all possible regressions that illustrates the Cai et al. (2009) approach, click here. .
5. For R compatible (chap17R.txt) and Mplus-compatible (Chap17M.txt) data files for the numerical example in Chapter 17, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id support coping pos riskHIV sevHIV depress supportb copingb posb riskHIVb sevHIVb depressb treat adhere
CHAPTER 18: INTRODUCTION TO MODERATION ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a webpage by David Kenny on moderation, click here.
3. For on-line calculators by Kris Preacher for probing moderation/interaction effects in multiple regression, multilevel models, and latent class analysis, click here.
4. For a webpage by Jorge Cortina with several on-line calculators related to moderation, click here.
CHAPTER 25: GROUP ADMINISTERED INTERVENTIONS AND CLUSTER DESIGNS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a paper that presents a clear and non-technical introduction to standard error corrections due to clustering, click here
3. For the NIH Pragmatic Trials Collaboratory for clustered randomized trials, click here.
4. For the Amazon page for the classic book on cluster randomized trials (2017) by Hayes and Moulton, click here.
5. For a paper on the use of MSEM to analyze a cluster randomized trial with binary outcomes, click here. For the Mplus code used in the paper, click here.
6. For a paper that applies MSEM to three level data by Kris Preacher, click here. For supplemental materials for examples 1, 2 and 3 in the paper, look at the last entry on Kris' webpage located here.
7. For Mplus references on MSEM at the Mplus website, click here.
8. For a useful video course (on demand) on MSEM, click here.
9. For a paper that describes MSEM approaches to the analysis of partially nested designs, click here. This paper is somewhat dated and relies on maximum likelihood as opposed to Bayesian methods for MSEM. For the supplement to this paper (which is a bit hard to find) with Mplus code for the examples in the article, click here.
10. For a document that shows you how to calculate an intraclass correlation in MSEM using Mplus with Bayesian estimation, click here
11. For an annotated output of the vigorous exercise example using Bayesiam MSEM, click here.
12. For a useful article on muliple group multilevel analyses, click here.
13. For an Mplus compatible data file for the vigorous exercise example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): mvpa peers advant treat school
14. For an Mplus compatible data file for the group administered face mask example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): intent norm att treat clus
15. For a dcument that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 26: MISSING DATA
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on how to apply FIML for missing data in lavaan, click here.
3. For the Part I video of the Nebraska Methodology Workshop (2015) by Craig Enders on missing data analysis, click here.
4. For the Part II video of the Nebraska Methodology Workshop (2015) by Craig Enders on missing data analysis, click here.
5. For the Mplus webpage with links to articles on missing data analysis in SEM, click here.
6. For Mplus webnotes with some data sets and material related to missing data analysis in SEM, click here.
7. For an introduction to 5 packages in R for addressing missing data, click here.
8. For a youtube video on using the R software MICE, click here.
9. For a youtube video applying selection modeling to missing data that are MNAR, click here.
10. For the BLIMP website and videos by Craig Enders for a wide range of missing data analyses, click here.
CHAPTER 27: INTENT TO TREAT AND PER PROTOCOL MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a description of and critique of intent-to-treat analyysis by Jerry Dallal of Tufts University, click here. For a low cost ($3) copy of the e-book in which this apperars, called The Little Handbook of Statistical Practice, click here.
3. For a document on applying IPTW weighting to binary outcomes for per protocol analyses, click here.
4. For a document on using CACE modeling in Mplus for in conjunction with missing data or treatment dropouts, click here.
5. For a free, downloadable book Causal Inference: What if by Hernán and Robins that includes material on per protocol analyses in randomized trials, click here. The book uses a potential outcomes and counterfactual approach.
6. To download the Mplus compatible data file for the direct covariate per protocol analysis in Chapter 27, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob
7. To downlaod the Mplus compatible data file for the IPTW per protocol analysis in Chapter 27, click here. Use this data for the doubly robust version of IPTW analysis as well. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob wght stabwght
8. To download the Mplus compatible data file for the CACE example in Chapter 27, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob;
CHAPTER 28: SAMPLE SIZE CONSIDERATIONS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For Kris Preacher's web-based programs on power analysis in SEM and assorted other issues, click here.
3. For a document on localized simulations for sample size decision making entitled Simulation Variants in Mplus, click here.
4. For a variety of on-line power analysis tools for conventional tests, click here and here.
5. For the R library semTools that has many useful power analysis utilities, click here.
6. For the popular free power software G*Power, click here.
7. For Paul Tremblay's website with Mplus power analysis simulation examples, click here.
8. For the classic Mplus paper on power analysis simulations by Linda Muthen et al. (2002), click here.
9. For an introduction to power analysis by the UCLA group, click here.
10. For a youtube video on how to program a localized simulation for power analysis in Mplus using parameter estimates from an existing data base (rather than a priori specifying your own parameter values), click here.
11. For two youtube videos (part 1 and part 2) on how to program a localized simulation in Mplus, click here and here.
.
T
1. For answers to frequently asked questions for this chapter topic, click here.
CHAPTER 2: CONCEPTUAL FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For drawing influence diagrams, a flexible program is Microsoft Visio, which is available through many universities. It also is available at a reasonable cost through third party vendors, like Brytesoft. Click below for a tutorial on how to use this program and a description of tools for use in Visio.
Read a short tutorial on how to use Visio.
Download templates and vss files discussed in the tutorial (this is a zip file that, when opened, will make 4 files available). You may have to adjust your Trust Settings in Visio to use these files - see the tutorial.
Another viable graphics program for influence diagrams that is free is app.diagrams. Click here for it.
3. For use on your computer without an internet connection, you can download the program that the daggity.net website uses to identify confounds, independence relations, and instrumental variables here. A pdf manual for the program is here.
a. A free, self-paced course on using DAGs in causal analysis is offered by Harvard University here.
b. DAGs conceptualize latent variables somewhat differently than SEM; latent variables are unobserved variables with no indicators whereas in SEM, latent variables often (but do not always) have reference indicators. For a technical discussion of the role of latent variables in DAGs, click here.
c. A good introductory treatment of DAGs is in Pearl, Glymour, and Jewell (2016).
4. For a copy of the Cinelli, Forney and Pearl (2022) paper on good and bad controls and arguments against atheoretical partialling, click here.
5. For a document on collider bias and mediation analysis, click here
CHAPTER 3: MEASUREMENT FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a brief overview of key concepts related to scaling theory for the construction of multi-item scales, click here.
3. For a primer on testing for unidimensionality of a multi-item scale, click here.
4. For a primer on strategies for addressing measurement error for single indicator SEM models, click here.
5. For a primer on testing measurement invariance, click here. To download the Mplus data file used in the primer, click here. Use the following names for the variables in this order in the Mplus syntax: id d1 d2 d3 d4 d5 d6 adhere adhere2 dfemale income ethnic.
6. For an open access article on a brief social desirability scale (that includes the scale itself), click here. I sometimes adapt the scale based on my study context and population.
7. For the chapter from Jaccard and Jacoby (2020) that describes application of the framework focused on the processes of comprehension, judgment, and response translation to measure construction, click here.
CHAPTER 4: METHODOLOGICAL FUNDAMENTALS FOR RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For an Amazon link to used books of the classic book on experimental and quasi-experimental design by Campbell and Stanley, click here. This is well worth a read.
3. For an updated (but still somewhat old) version of the Campbell and Stanley book (a lengthy chapter by Sadish, Campbell and Stanley), click here. It is downloadable from the internet.
4. For a more recent and excellent text on experimental design and quasi experimental design by Charles Reichardt titled "Quasi-Experimentation: A Guide to Design and Analysis", click here. There is a wealth of information about experimental design per se.
5. For a youtube introductory video on common clincial trial designs, click here.
CHAPTER 5: STATISTICAL FUNDAMENTALS FOR RETs: REGRESION
1. For answers to frequently asked questions for this chapter topic, click here.
2. There are a large number of books on the different regression methods discussed in Chapter 5 making recommendations challenging. Sage publishes a series of relatively introductory, inexpensive monographs, referred to by some as the "little green books" that cover many statistical topics, including regression. For a list of the regression books in the Sage series, click here. The SAGE regression webpage includes links in the "search by category" box for other little green books.
3. A readable competing series of introductory monographs (called "Blue Books") is offered by David Garson. Many of these monographs are available on Amazon as ebooks and they are free. For a list of these monographs, click here. Be sure to focus on from this list those monographs that are in the Blue Book series.
4. Guilford Publications offers high quality, readable and reasonably priced paperbacks on a range of regression related topics as part of their series "Methodology and Statistics". For a link to the books in this series, click here. Lot's of great books in this series, so it is worth checking out.
CHAPTER 6: STATISTICAL FUNDAMENTALS FOR RETs: ADVANCED TOPICS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a primer on smoothers, click here.
3. For a primer on quantile regression, click here.
4. For an introductory youtube video on instrumental variables by the Nobel Prize Laureate Joshua Angrist, click here. This is one video in a series of videos on econometrics.
5. For an introductory youtube video on margins of error, click here.
6. For an introductory youtube video on centering, click here.
7. Rand Wilcox at USC is an expert on robust analyses and has written numerous books on them that range from clearly written introductory treatments to more advanced treatments. I highly recommend his books and articles. For his webpage, click here. Be sure to click on the link to his CV on this webpage.
8. For an introduction to bootstrapping and two useful introductory videos to bootstrapping, click here.
CHAPTER 7: STATISTICAL FUNDAMENTALS FOR RETs: STRUCTURAL EQUATION MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a video of the SEM decomposition process described in Chapter 7, click here.
3. For a useful website by Dave Kenny on many SEM topics, click here.
4. To access a popular SEM blog called SEMNET, click here. Two other potentially useful blogs for SEM advice include the blog for the lavaan Google group (click here) and a blog called CrossValidated that includes more general statistical questions and answers (click here).
5. For the webpage of Rex Kline's introductory book on SEM, click here.
6. For the webpage of Tim Brown's book on confirmatory factor analysis, click here.
7. For the webpage of Rick Hoyle's Handbook of SEM, click here.
8. For a document that describes SEM estimation algorithms in Mplus, click here.
9. For a youtube video on fit indices in SEM, click here.
10. Patrick Curran of the University of North Carlolina provides a free 6 you tube video series on SEM as part of CenterStat. To access these videos (SEM Episodes 1 to 6), click here. There are multiple other videos on this web page you may find of interest, such as a nine video treatment of growth curve modeling and some videos (by Dan Bauer) on latent class mixture models
and a seven video series on multiple regression.
11. Mikko Rönkkö has published on youtube over 200 videos on different facets of SEM. To go to his youtube page for the videos, click here.
CHAPTER 8: STATISTICAL FUNDAMENTALS FOR RETs: NON-TRADITIONAL SEM
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a pdf of the classic book on Bayesian Data Analysis by Gelman et al., click here.
3. For a website on piecewise SEM and an R package for it, click here.
4. For a document that describes how to apply quantile regression to RETs using limited information SEM, click here.
5. For a youtube video about Bollen's model implied instrumental variable approach, click here.
7. For a useful powerpoint on the use of Bollen's model implied instrumental variable approach, click here.
8. For Judea Pearl's homepage with many resources for Structural Causal Modeling, click here.
9. For a primer on how to locate the asymptotic covariance matrix in Mplus, lavaan, MIIVsem, STATA, R, and SPSS output for the calculation of Monte Carlo confidence intervals, click here.
CHAPTER 9: MEDIATION ANALYSIS IN RETs: BASIC APPROACHES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a consensus panel's recommendations on what to report for mediation analyses, click here.
3. For the web page on mediation on the Mplus website, click here.
4. For Tyler VanderWeele's homepage with useful videos on casual mediation frameworks, click here.
CHAPTER 10: EVALUATING EFFECT SIZES IN RETs
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a web site to calculate traditional standardized effect sizes, click here. The probability of exceptions to the rule index, expressed as a percentage, can be calculated by subtracting the CLES (common language effect size) index on this website from 100. To convert one effect size index to another effect size index, see entry number 14 called Transformation of the effect sizes d, r, f, odds ratio, eta squared, and CLES. The site also has a nice section on testing the significance of correlations that implements many tests of differences between correlations for independent and dependent groups.
3. For an interactive, graphical illustration of the relationship of Cohen's d to overlap indices and the probability of exceptions to the rule, click here. This website calls the common language effect size the probability of superiority and expresses it in percentage form. Subtracting it from 100 yields the probability of exceptions to the rule, expressed as a percentage. The website includes an index of the number needed to treat (NNT) for continuous outcomes using a cutoff score to dichotomize the continuous outcome, which, in my opinion, is dubious because it falsely dichotomizes a continuous construct. The website allows you to alter the default value of the cutoff (see the formulas section of the website). Having said that, if you work with a substantively defined, theoretically meaningful cutoff value, this is a reasonable tool to graphically explore the relationship between d and NNT with cutoffs.
4. For a primer on how to conduct hierarchical regression analysis in Mplus in order to obtain indices of unique explained variance using robust estimators such as MLR and how to use MM regression to isolate outlier resistant indices of incremental explained variance, click here.
CHAPTER 11: MEDIATION ANALYSIS WITH CONTINUOUS OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a video that talks you through Mplus syntax for the numerical example in Chapter 11, click here.
3. For a document that describes preliminary analyses for the social phobia example, click here.
4. A misunderstood warning message that sometimes occurs in Mplus reads "THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION." The warning appears in the output of the example for Chapter 11. Sometimes this warning requires model modification and sometimes it can reasonably be ignored. For an Mplus document that describes the technical bases of the message, click here. The warning can be safely ignored in the example for Chapter 11. In my opinion, the Mplus document sometimes creates the impression that corrective action might be needed in certain cases where it probably is not, but the presence of the message in and of itself should cause you to reflect on the matter and ensure you are comfortable moving forward, perhaps after looking at further diagnostics.
5. For a document on measurement non-invariance tests for the latent social phobia indicators in the chapter example, click here.
6. For a document that describes how to calculate standardized effect sizes for the chapter example, click here.
7. For a document that shows how to use Mplus to perform effect decompositions, click here.
8. For a document that shows how to calculate or obtain from Mplus classic omnibus mediation tests and interventional indirect effects for the social phobia example, click here.
9. For a document that describes how to conduct sensitivity tests in full information SEM, click here.
10. For a document that describes how to deal with unmeasured confounds and correlated distrubances using the social phobia example, click here.
11. For a document that shows how to get basic descriptive statistics in Mplus using the BASIC option, click here.
12. For a document that describes tests for competing models to the social phobia model, click here.
13. For a document that shows how to adjust single indicators for measurement error in the chapter example, click here.
14. For an annotated output of the chapter example using traditional FISEM, click here.
15. For the DAGitty code to identify independence tests for the social phobia example, click here.
16. For R-compatible (chap11R.txt) and Mplus-compatible (chap11M.txt) data files for the social phobia example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id cr1 spai1 spin1 cr3 spai3 spin3 negapp2 pskills2 extern2 negapp1 pskills1 extern1 hyper sex treat
17. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 12: MEDIATION ANALYSIS WITH BINARY OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a document that describes preliminary analyses for the parent communication example, click here
3. For the DAGitty code to identify independence tests for the parent communication example, click here.
4. . For a document that shows how to calculate or obtain from Mplus the classic omnibus mediation tests for the parent communication example, click here. This includes material on omnibus tests for the causal mediation framework.
5. For a document that shows how to calculate margins of error and confidence intervals for indices of relative risk and the number needed to treat using Mplus, click here.
6. For programming the chapter example using logit instead of probit modeling, click here.
7. For annotated output of the chapter example using traditional probit-based modeling, click here.
8. For an informative video on the shortcomings of logistic regression by Edward Norton, a leading economist, click here. This video is very slow to download so be patient. Also check out the handouts and transcript for the video on that same webpage.
9. For the classic article by Mood critiquing logistic regression, click here. For an article by Gomila comparing logistic regression and the linear probability model, click here and for a critque of logistic regression by Norton, click here.
10. For R-compatible (chap12R.txt) and Mplus-compatible (chap12M.txt) data files for the parent communication example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id com3 pa2 pk2 pe2 cq1 pa1 pk1 pe1 t bs1
11. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 13: MEDIATION ANALYSIS WITH ORDINAL AND NOMINAL OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on ordinal regression, click here.
3. For a useful youtube video on how to compute marginal effects for predictors in ordinal regression using STATA, click here.
4. For a youtube video on working with ordinal level predictors, click here. In the video, I do not agree with the assertion that all of the coefficients for staircase based dummy variables need to be statistically signficant. I believe that as long as at least one is statistically signficant that this suggests something is going on. But aside from this quibble, the video is pretty good.
5. For annotated output of the chapter example using traditional probit-based analysis, click here.
6. For R-compatible (chap13R.txt) and Mplus-compatible (chap13M.txt) data files for the clinician training example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): ID GA2 TA2 BD2 CE1 CIS1 TREAT IMP3 y234 y34 y4 cCE1 cCIS1.
7. For a document called Supplenetary Total Efffects Analysis for Ordinal and Multinomial Modeling that provides a nonparametric method for analyzing total effects in ordinal or multinomial regression, click here.
8. For a document that illustrates how to write up the results for the chapter example, click here.
CHAPTER 14: MEDIATION ANALYSIS WITH COUNT OUTCOMES
1. For answers to frequently asked questions for this chapter topic, click here.
CHAPTER 15: NON-LINEAR AND SPECIALIZED MODELING IN MEDIATION ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on quadratic regression, click here.
3. For a youtube video on regression splines, click here.
4. For an introductory book chapter by Jaccard on how to build math models using traditional non-linear regression approaches (and more), click here.
5. For two useful youtube videos on Bayesian Additive Regression Trees (BART), click here and here. These videos develop the mathematical details of BART in more detail than I do in my chapter but are still somewhat conceptual in focus.
6. For introductory youtube videos on generalized additive models by Noam Ross, click here, here and here. For an additional workshop video by Noam, click here. For a free inteactive course on GAMs by Noam, click here.
7. For a brief primer on cluster plots when a cluster analysis focuses on more than two target variables, click here.
8. For an introductory youtube video on principal component analysis, click here
9. For an introductory youtube video on discriminant analysis, click here
10. The R program I use for variants of trimmed k-means analysis is tclust. It is dated. A newer version of tclust that offers more flexibility and features requires special downloading steps than is typical in R. For instructions, click here. The program I use on my website is the older version. You only need to download the newer version if you are going to do more advanced R programming on your own independent of my program. Otherwise you can just wait until the newer version makes it to CRAN.
11. For R-compatible (quadraticR.txt) and Mplus-compatible (quadraticM.txt) data files for the quadratic RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): treat m1 m2 m3 m1m1 y v1 v2 v3 v4. Exclude variables v1-v4 with the USEVARIABLES command within Mplus.
12. For R-compatible (splineR.txt) and Mplus-compatible (splineM.txt) data files for the spline modeling RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id m1 m2 y treat covm1 covm2 covy ccovy.
13. For R-compatible (nonlinearR.txt) and Mplus-compatible (nonlinearR.txt) data files for the traditional non-linear regression RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id y m1 m2 treat.
14. For R-compatible (bartdat.txt) data file for the BART example, click here.
15. For R-compatible (rcluster2R.txt) and Mplus-compatible (rcluster2M.txt) data files for the trimmed k-means robust cluster analysis RET example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id warmth control clus treat d1clus d2clus d3clus d4clus cov1 cov2 y. Exclude variables from the analysis with the USEVARIABLES command, as appropriate.
16. For a document that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 16: RETs WITH FOLLOW-UPS: LONGITUDINAL MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a useful article on lagged dependent variables, click here. For a useful discussion by Paul Allison on the use of lagged dependent variables in mixed models, click here.
3. For a document that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 17: MEDIATOR RELATIVE IMPORTANCE AND EXPLORATORY MEDIATOR ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a link to the relaimpo package in R for relative importance analysis, click here.
3. For a link to conducting dominance analysis using STATA, click here.
4. For a worked example of mediator selection using all possible regressions that illustrates the Cai et al. (2009) approach, click here. .
5. For R compatible (chap17R.txt) and Mplus-compatible (Chap17M.txt) data files for the numerical example in Chapter 17, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): id support coping pos riskHIV sevHIV depress supportb copingb posb riskHIVb sevHIVb depressb treat adhere
CHAPTER 18: INTRODUCTION TO MODERATION ANALYSIS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a webpage by David Kenny on moderation, click here.
3. For on-line calculators by Kris Preacher for probing moderation/interaction effects in multiple regression, multilevel models, and latent class analysis, click here.
4. For a webpage by Jorge Cortina with several on-line calculators related to moderation, click here.
CHAPTER 25: GROUP ADMINISTERED INTERVENTIONS AND CLUSTER DESIGNS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a paper that presents a clear and non-technical introduction to standard error corrections due to clustering, click here
3. For the NIH Pragmatic Trials Collaboratory for clustered randomized trials, click here.
4. For the Amazon page for the classic book on cluster randomized trials (2017) by Hayes and Moulton, click here.
5. For a paper on the use of MSEM to analyze a cluster randomized trial with binary outcomes, click here. For the Mplus code used in the paper, click here.
6. For a paper that applies MSEM to three level data by Kris Preacher, click here. For supplemental materials for examples 1, 2 and 3 in the paper, look at the last entry on Kris' webpage located here.
7. For Mplus references on MSEM at the Mplus website, click here.
8. For a useful video course (on demand) on MSEM, click here.
9. For a paper that describes MSEM approaches to the analysis of partially nested designs, click here. This paper is somewhat dated and relies on maximum likelihood as opposed to Bayesian methods for MSEM. For the supplement to this paper (which is a bit hard to find) with Mplus code for the examples in the article, click here.
10. For a document that shows you how to calculate an intraclass correlation in MSEM using Mplus with Bayesian estimation, click here
11. For an annotated output of the vigorous exercise example using Bayesiam MSEM, click here.
12. For a useful article on muliple group multilevel analyses, click here.
13. For an Mplus compatible data file for the vigorous exercise example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): mvpa peers advant treat school
14. For an Mplus compatible data file for the group administered face mask example, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): intent norm att treat clus
15. For a dcument that illustrates how to write up the results for the chapter examples, click here.
CHAPTER 26: MISSING DATA
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a youtube video on how to apply FIML for missing data in lavaan, click here.
3. For the Part I video of the Nebraska Methodology Workshop (2015) by Craig Enders on missing data analysis, click here.
4. For the Part II video of the Nebraska Methodology Workshop (2015) by Craig Enders on missing data analysis, click here.
5. For the Mplus webpage with links to articles on missing data analysis in SEM, click here.
6. For Mplus webnotes with some data sets and material related to missing data analysis in SEM, click here.
7. For an introduction to 5 packages in R for addressing missing data, click here.
8. For a youtube video on using the R software MICE, click here.
9. For a youtube video applying selection modeling to missing data that are MNAR, click here.
10. For the BLIMP website and videos by Craig Enders for a wide range of missing data analyses, click here.
CHAPTER 27: INTENT TO TREAT AND PER PROTOCOL MODELING
1. For answers to frequently asked questions for this chapter topic, click here.
2. For a description of and critique of intent-to-treat analyysis by Jerry Dallal of Tufts University, click here. For a low cost ($3) copy of the e-book in which this apperars, called The Little Handbook of Statistical Practice, click here.
3. For a document on applying IPTW weighting to binary outcomes for per protocol analyses, click here.
4. For a document on using CACE modeling in Mplus for in conjunction with missing data or treatment dropouts, click here.
5. For a free, downloadable book Causal Inference: What if by Hernán and Robins that includes material on per protocol analyses in randomized trials, click here. The book uses a potential outcomes and counterfactual approach.
6. To download the Mplus compatible data file for the direct covariate per protocol analysis in Chapter 27, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob
7. To downlaod the Mplus compatible data file for the IPTW per protocol analysis in Chapter 27, click here. Use this data for the doubly robust version of IPTW analysis as well. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob wght stabwght
8. To download the Mplus compatible data file for the CACE example in Chapter 27, click here. The names of the variables that should appear on the NAMES ARE command for the Mplus syntax are (in order): depress risk Tx depbase age motivate educ assert single econ nonwhite x10 c1 c2 tncomply depchang prob;
CHAPTER 28: SAMPLE SIZE CONSIDERATIONS
1. For answers to frequently asked questions for this chapter topic, click here.
2. For Kris Preacher's web-based programs on power analysis in SEM and assorted other issues, click here.
3. For a document on localized simulations for sample size decision making entitled Simulation Variants in Mplus, click here.
4. For a variety of on-line power analysis tools for conventional tests, click here and here.
5. For the R library semTools that has many useful power analysis utilities, click here.
6. For the popular free power software G*Power, click here.
7. For Paul Tremblay's website with Mplus power analysis simulation examples, click here.
8. For the classic Mplus paper on power analysis simulations by Linda Muthen et al. (2002), click here.
9. For an introduction to power analysis by the UCLA group, click here.
10. For a youtube video on how to program a localized simulation for power analysis in Mplus using parameter estimates from an existing data base (rather than a priori specifying your own parameter values), click here.
11. For two youtube videos (part 1 and part 2) on how to program a localized simulation in Mplus, click here and here.
.
T