Winter, Spring 2004
| access 2002 version course content at
SUSE Computing Lab in CERAS
LISREL SAS and HLM are
supposedly installed in CERAS computer lab.
Raudenbush, Stephen W., & Bryk, Anthony S. Hierarchical Linear Models. Applications and Data Analysis Methods. Sage, 2nd edition, 2002.
Kline Rex B. Principles and Practice of Structural Equation Modeling (1st ed.) 1998, Guilford Publications, Inc.
A partially knowledgable observer could describe this course by the buzz-words "LISREL and HLM" and that concise phrase is somewhat informative. A main objective is to take a serious look at some of these advanced (and heavily marketed) statistical procedures that have become widely used (for better or worse) in education and social science. The broader perspective is to start with the data analysis (and substantive) settings that these procedures purport to address (if not solve):
1. Analysis of Multilevel Data (e.g., kids within classrooms within
The point being that there is much much much more to these important topics than what is covered by LISREL and HLM (programs or writings) and the challenge of organizing this course is to weave in the larger issues.
events of interest
|Bringing Evidence-Driven Progress
main report November 2002 US DOE press release December 2003 confab, "what works"
Educational Policy Example: Charter Schools
Educational Policy Example: Teacher Credentialling
State of art example
Justin Tobias, UC Irvine: Bayesian modeling of school effects using hierarchical models with smoothing priors.
Key Resource for Causal Inference: Winship's repository Counterfactual
Causal Analysis in Sociology
for Multilevel Modelling (H Goldstein)
for mixed-effects models
MIXED and NLMIXED
James L. Arbuckle
Social Sciences Data Resources
Jan 6. Course Introduction.
Current Event: Coffee and diabetes studies
Science Daily: Long-Term Coffee Consumption Significantly Reduces Type 2 Diabetes Risk Coffee: A new miracle drug for diabetics?
2. Jan 13. Evidence
driven education readings.
Adventure #1. Multilevel school data taken from the MlWin
The sequence of the
variables and the coding are as follows:
For the data in Adventure 1, use normexam as outcome and standirt (pretest)
Evidence driven education readings.
Data Adventure #1 discussion
Derivation Contentual Effects Relations (DCD)
Current Event : Vitamins and Alzheimer's
Science Daily: Vitamin Supplement Use May Reduce Effects Of Alzheimer's Disease Pravda: Vitamins E & C to fight Alzheimer's
Analysis of Covariance and Comparing Regressions.
Equivalence to HLM (High School and Beyond) Analyses; HLM example
Introduction to Path Analysis and Structural Equation Models Ed257 path example
Text Readings Kline Chap. 3, Chap 5 (esp 5.8, 5.10)
Current Event: Sleep and Math Performance
BBC: Sleep 'can increase brain power' Nature: Sleep boosts lateral thinking
Multiple Regression Parameters (Mosteller & Tukey exs)
Measurement error and regression; multiple regression estimates via normal eqs (path anal);
Path analysis examples (Freedman notes and paper);
Intro to structural equation models (Allison notes); notation and estimation handout;
Text Readings Kline Chap. 7, Chap 8
Adventure # 2
[for Feb 10 class
Data Adventure # 3 [for Feb. 10 class] (From Paul Allison course notes)
class 1.00 famsize -.33 1.00 ability .39 -.33 1.00 esteem .14 -.14 .19 1.00 achieve .43 -.28 .67 .22 1.00
Do the indicated path analysis and interpret.
Feb 17 and Feb 24
Discussion: critique of structural equation models
Current event: Diabetes and child obesity, Causal-correlational farce?
California Center for Public Health Advocacy report
b. Holland-Rubin models for comparative experiments (causal inference)
Causal Inference, Path Analysis, and Recursive Structural Equations Models Paul W. Holland Sociological Methodology, Vol. 18. (1988), pp. 449-484.
Abstract Rubin's model for causal inference in experiments and observational studies is enlarged to analyze the problem of "causes causing causes" and is compared to
path analysis and recursive structural equations models. A special quasi-experimental design, the encouragement design, is used to give concreteness to the discussion by
focusing on the simplest problem that involves both direct and indirect causation. It is shown that Rubin's model extends easily to this situation and specifies conditions
under which the parameters of path analysis and recursive structural equations models have causal interpretations.
NOTE: this is a JSTOR link so it requires you to be on a Stanford IP machine (i.e. campus or campus dial-up or use proxy server)
Related technical reading
Statistics and Causal Inference, Paul W. Holland pp. 945-960 JASA 1986, another JSTOR link
Commentaries Donald Rubin, David Cox
Dead week mtg
Causal current event, folic acid and heart attacks
Student presentations, research papers
|SPRING QUARTER 2004|
Causal current event: stents and heart attacks , value of experiments
intro and organization
Causal current event: music downloads and CD sales
instrumental variables analysis, Freedman stat151 book, chap 8
HLM review and discussion: B&R text chap 1-5.
current event: TV
Watching TV may hurt toddlers' attention spans Toddlers' TV viewing linked to attention deficit
April 13. AERA Week.
Discussion student Research 4PM
Continue Instrumental variables, Freedman Chap 8.
Three-level HLM examples, B&R text Chap 8, SSI site
Student research projects
Non-linear Multilevel Models (counts and proportions)
RB text Chap 10 (logistic and Poisson link functions)
nlme (Bates-Pinhero book) Mixed-Effects Models in Practice lme for SAS PROC MIXED Users user guide
SAS NLMIXED sugi papers see SAS v9 docs on Ceras machines
Russell D. Wolfinger, SAS Institute Inc., Fitting Nonlinear Mixed Models with the New NLMIXED Procedure
non-Compliance in experiments and trials. Guest lecturer Booil Jo, Dept of Psychiatry
readings: Statistical Power in Randomized Intervention Studies With Noncompliance
Model misspecification sensitivity analysis in estimating causal effects of interventions with non-compliance
Estimation of Intrevention Effects with Noncompliance
18. May 18.
Principal Stratification Approach to Broken Randomized Experiments: A Case Study of School Choice Vouchers in New York City [pdf on bb.stanford.edu, Ed260]
Causal current event
Vigorous Exercise May Slow Women's Bone Loss Study: Earlier bone-booster use may limit osteoporosis
Continuing Topics. Followup on topics RB Chap 6, 10; selected topics RB Ch 11.
June 1. Dead Week Meeting
Student Research Projects
Causal Current event: Study: Driving longer means larger waists