Psy 613 Data Analysis III:

Multivariate Techniques

 

Spring 2008

Tue, Thu 10:00-11:20, 189 PLC

Lab: Fri 10:00-12:00, 180 Straub

 

Instructor: Bertram Malle

Straub 305, 346-0475

bfmalle@uoregon.edu

Office hours: Thu 12 p.m. & by apptmt.

Teaching Assistants:

Amber Thalmeyer

Laura Kaehler

Straub 351, 346-4930

Straub 383, 346-3936.

athalmay@uoregon.edu

lkaehler@uoregon.edu

Tue 12-1 p.m., Wed 3-4 p.m.

Tue 2:00-4:00 p.m.

 

Overview. This course covers the basic multivariate techniques that are currently used in psychology and other social and life sciences. These include principal components and factor analysis, multivariate analysis of variance, discriminant function analysis, logistic regression, and log-linear analysis. Other techniques, such as structural equation modeling, hierarchical linear modeling, canonical correlation, or multidimensional scaling, are not covered in detail but are briefly introduced at the end of the term.

The learning goal for students is to have a conceptual and statistical understanding of each technique, be able to apply the correct technique to any given data set, properly interpret the output of statistical computer packages (primarily SPSS), and understand and critique scientific papers that use these techniques.

Even though much work will go into learning to conduct the various analyses, a primary goal is to gain a conceptual understanding of each technique, which refers to its function and capacity to answer scientific questions (not to achieve p values). Each technique is a tool that can immensely assist you in understanding your data and testing your hypotheses; but to apply each tool correctly you need to understand both your data and the nature of these tools.

The level of understanding you will reach in this course lies one step deeper than the level at which you will operate when using the techniques in years to come. During this term you will learn to think about some mathematical underpinnings of the techniques, and this step will give you a deeper understanding of the workings of each technique. Even if you later forget the details of these underpinnings, you will not forget the conceptual implications they have. All mathematics introduced in this course is meant to serve your conceptual understanding.

Because most or all techniques will be new to you, we need to create maximal familiarity through redundancy. You are expected to read the relevant textbook chapters before each lecture. The lectures themselves will introduce the material from a conceptual and statistical perspective, with directions to computer analysis. The lab on Friday will briefly review the lectures and then introduce relevant SPSS commands to analyze actual data using the statistical techniques. The lab will also discuss the current homework and review past weekÕs homework. In the homeworks, you will run analyses on provided data sets and write up the results, culminating in a one-page summary that resembles the report in an empirical journal article. Finally, the exams will ask you to show your command over the conceptual underpinnings of each technique.

 

Grading.  Your grade will be based on 9 homeworks (100 pts each) and two take-home exams (300 points each), for a total of 1500 points. Tentative grade cut-offs are 1350 (A) and 1200 (B). As mentioned, the homeworks consist of data retrieval, analysis, interpretation and write-up of results, all done electronically. They will be graded by the teaching assistants using explicit criteria that I provide. The teaching assistants have taken this course previously and performed at a superior level. If anyone has concerns about their work being graded by a fellow graduate student, please approach me. The exams (web-posted May 3 and June 5) consist of five to six questions on the functional character of techniques and your understanding of central concepts.  I grade the first exam; TAs help with a few questions on the final exam, again following an explicit grading scheme.

 

Readings: Because there is no all-around good, complete, and affordable textbook of multivariate statistics, I have selected readings from various textbooks and other sources for this class (all available on Blackboard).  You are advised to at least skim the relevant chapters before class to familiarize yourself with the terminology.  After class you should read the chapter in detail to deepen you understanding, which will speed up your homework and prepare you well for the exams. 

            For those of you who would like to own a source book, I recommend Tabachnick and Fidell (T & F), Multivariate Statistics (2006, 5th ed.).  You are not required to buy T & F, but if you do, you must still read the other  required chapters, because for some topics T & F just donÕt provide the background you need.

 

Resources:  Extensive lecture handouts are posted on the Blackboard page and on a course web page (www.uoregon.edu/~bfmalle/613.html). Each lecture is also audio recorded and made available on Blackboard.  Before each lecture, last yearÕs material can be inspected; after ThursdayÕs lecture, the weekÕs material will be updated where appropriate. The Blackboard page also contains the electronic reading library and additional resource material.

            I assume that you have reliable access to SPSS statistical software and that you can without a problem exchange electronic documents with your TAs.

 

Homework: Each Wednesday, a homework assignment will be posted on the web and is due the following Thursday, 10 a.m.  Given the long time to complete the homework, extension will be given only in extraordinary circumstances and only before the due date.  Points will be subtracted for lateness.

            Analyses relevant to the homework will be practiced during lab hours on Friday (10:00am-12:00 in Straub 180). Typically the assignment contains a data set and several instructions to analyze it. Your homework is complete if you run all required analyses, edit the output files down to the essential information (there will typically be a page limit), and annotate the output to demonstrate your understanding of what SPSS was doing.  Finally, you write up a one-page summary of the results as you would for an empirical journal article.  The annotated outputs and the written results must be sent to your TA via email by Thursday, 10 a.m.  For the written results, you can take Tabachnick & FidellÕs result summaries or two sample summaries on the course web page as models (http://darkwing.uoregon.edu/~bfmalle/613/Summaries.html). Be sure to write clearly, concisely, and meaningfully.

            For all of your homeworks I strongly recommend that you have the SPSS Syntax command reference document handy (available in the SPSS Help pull-down menu and on Blackboard). Studying it can save you many hours of SPSS frustration.  If you are stuck with an SPSS problem for more than 10 minutes, contact me or your TAs to help you solve it.

Topics and Readings

The readings in this class are selected chapters from various statistics textbooks, SPSS handbooks, and a few journal articles. In addition, I have listed the relevant Tabachnick & Fidell (2006) chapters where available. In some cases, a Tabachnick & Fidell chapter is required reading.  Required readings are printed in blue font below.

            All reading materials are available as pdf files on Blackboard. If you need to check out paper documents for copying, please contact me. 

 

Textbook authors cited below:

Cliff, N. (1987).  Analyzing multivariate data. Harcourt Brace Jovanovich: San Diego.

Dillon, W. R., & Goldstein, M. (1984). Multivariate Analysis: Methods and Applications. Wiley: New York.

Tabachnick, B. G., & Fidell, L. S. (2006).  Using multivariate statistics (5th ed.). New York: HarperCollins.

 

L 1: Tu Apr 1  The multivariate approach: Introduction and overview

È Dillon & Goldstein (1984). Overview of multivariate techniques (pp. 19-22) 

È Tabachnick & Fidell (2006). ch. 1 + ch. 2. 

Optional

SPSS Syntax command reference

 

L 2: Thu Apr 3 Data screening and exploratory data analysis

È Tabachnick & Fidell (2006). ch. 4

È Cohen, J. (1990). Things I have learned (so far).  American  Psychologist, 45, 1304-1312.

Optional

SPSS EXAMINE complete chapter

SPSS EXAMINE syntax reference

 

        Fri Apr 4  Lab 1: Exploratory data analysis

 

L 3: Tue Apr 8  Matrix algebra (Introduction)

È [very basic, slow pace, for beginners] Cliff, N. (1987). Elements of matrix algebra for statistical applications (ch. 1), Vectors (ch. 3).

È [faster pace, for intermediates] Dillon & Goldstein (1984). Vector and matrix operations and selected statistical concepts (pp. 521-539).

Optional

Tabachnick & Fidell (2006). Appendix A

 

L 4: Thu Apr 10 Matrix algebrafor statistics

È [faster pace] Dillon & Goldstein (1984). Statistical concepts and vector and matrix operations (pp. 6-18).

È [slower pace, thorough] Cliff, N. (1987). Statistical formulas in matrix form (ch. 2), Variances and covariances of linear combinations (ch. 4), The inverse (ch. 5).

Optional

Handout on determinants and inverses [advanced]

 

         Fri Apr 11  Lab 2: Matrix concepts

 

L 5: Tue Apr 15 Test Theory and Item Analysis

È http://en.wikipedia.org/wiki/Classical_test_theory

È SPSS Base System UserÕs Guide. ch. 26 (Procedure RELIABILITY)

Optional

[very thorough explanation of reliability] Nunnally, J. C. (1967). Theory of measurement error. In  J. C. Nunnally, Psychometric theory (ch. 6,  pp. 172-205). New York: McGraw-Hill.

Messick, S. (1995). Validity of psychological assessment: Validation of inferences from personsÕ responses and performances as scientific inquiry into score meaning. American Psychologist, 50, 741-749.

 

L 6: Thu Apr 17 Principal components analysis (PCA)

È Dillon & Goldstein (1984). Principal components analysis (ch. 2, pp. 23-39, 47-52)

È SPSS Base System UserÕs Guide. ch. 21 (Procedure FACTOR)

Optional

Cliff (1987). Components and principal components of variables (ch. 13)

Tabachnick & Fidell (2006). ch. 13

 

        Fri Apr 18  Lab 3: PCA

 

L 7: Tue Apr 22 PCA extensions

 

L 8: Thu Apr 24 Factor analysis (FA)

È Dillon & Goldstein (1984). Factor Analysis (ch. 3)

È (review from Apr 14) SPSS Base System UserÕs Guide. ch. 21 (Procedure FACTOR)

Optional

Cliff (1987). The common factor model (ch. 15)[ challenging but thorough.]

Gorsuch, R. L. (1997). Exploratory factor analysis: Its role in item analysis. Journal of personality assessment, 68, 532-560.]

Russell, D. W. (2002). In search of underlying dimensions: The use (and abuse) of factor analysis in Personality and Social Psychology Bulletin. Personality and Social Psychology Bulletin, 28,, 1629-1646.

 

         Fri Apr 25  Lab 4: Factor Analysis

 

L 9: Tue Apr 29 (M)ANOVA preludes: Interactions, orthogonality

È Rosnow, R. L., & Rosenthal, R. (1989). Statistical procedures and the justification of knowledge in psychological science.  American Psychologist, 44, 1276-1284.

È Rosnow, R. L., & Rosenthal, R. (1989). Definition and interpretation of interaction effects. Psychological Bulletin, 105, 143-146.  [make sure you read this--it will prevent you from making one of the most wide-spread statistical errors in psychology!]

 

L 10: Thu May 1 Significance testing and meta-analysis

È Schmidt, F. (1996). Statistical significance testing and cumulative knowledge in psychology: Implications for training of researchers.  Psychological Methods, 1, 115-129.

È DeCoster, J. (2004).  Meta-analysis.  In Kempf-Leonard, K. (Ed.),  The Encyclopedia of Social Measurement.  San Diego, CA: Academic Press.

Optional

Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183-192.

Dunlap, W. P., Cortina, J. M., Vaslow, J. B., & Burke, M. J. (1996). Meta-analysis of experiments with matched groups or repeated measures designs.  Psychological Methods, 1, 170-177.

 

Midterm take-home exam posted Thu May 1

           Fri May 2 Lab 5: Meta-analysis

 

L 11: Tue May 6 General multivariate analysis of variance (Manova)

È Tabachnick & Fidell (2006). ch. 9

È SPSS Advanced Statistics UserÕs Guide. ch. 3 (Procedure MANOVA)

È SPSS Manual: Appendix B–Categorical variable coding schemes.  

Optional

SPSS Keywords: Interpreting MANOVA parameter estimates.

 

Midterm take-home exam due Thu May 8, 4:00 p.m.

 

L 12: Thu May 8 Discriminant function analysis (DFA)

Dillon & Goldstein? È Tabachnick & Fidell (2006). ch. 11

È SPSS Advanced Statistics UserÕs Guide. ch. 1 (Procedure DISCRIMINANT)

Optional

Tabachnick & Fidell (2006). ch. 8

 

           Fri May 9  Lab 6: General Manova and DFA

 

L 13: Tue May 13 Univariate vs. multivariate repeated measures analysis of variance

È SPSS Advanced Statistics UserÕs Guide. ch. 4 (More on Procedure MANOVA)

 

L 14: Thu May 15 Multifactor multivariate repeated measures analysis of variance (profile analysis)

È Tabachnick & Fidell (2006). ch. 10 (pp. 441-483; pp. 503-505)

 

           Fri May 16  Lab 7: Repeated-measure analysis

 

L 15: Tue May 20 (M)Anova expansions (doubly multivariate designs; simple effects)

È SPSS handout on simple effects

È SPSS handout on doubly multivariate analysis

Optional

Tabachnick & Fidell (2006). ch. 3

 

L 16: Thu May 22 Logistic Regression

È Tabachnick & Fidell (2006). ch. 12

È SPSS Advanced Statistics UserÕs Guide. ch. 2 (LOGISTIC REGRESSION)

 

           Fri May 23  Lab 8: Logistic Regression

 

L 17: Tue May 27 From c2 to log-linear analysis

Other?È Tabachnick & Fidell (2006). ch. 7

 

L 18: Thu May 29 Log-linear and logit analysis

È SPSS Advanced Statistics UserÕs Guide. ch. 5 (Procedure HILOGLINEAR) and ch. 6 (Procedure LOGLINEAR)

 

           Fri 5/30  Lab 9: Log-linear analysis

 

L 19: Tue 6/3 Putting it all together: The multivariate tool box

Optional

Tabachnick & Fidell (2006). ch. 15, 2

 

Final take-home exam posted Tue Jun 3

 

L 20: Thu 6/5 Luxury tools: Canonical correlation, covariance structure analysis;

cluster analysis; multidimensional scaling.

Optional

Tabachnick & Fidell (2006). ch. 14

 

Fri 6/6  No lab

 

Final take-home exam due Tue Jun 10, 12noon