EDUCATION 257      Winter-Spring 2005

David Rogosa (, e314)

note: NWK chapter cites to fourth ed.; corresponding ver5 added alongside

I. Design and Analysis of Comparative Studies (Experiments)

A. Introduction and review. Factorial Designs
1. Comparing group outcomes on a single classification: One-way analysis of variance
2. Multiple comparisons in one-way anova
3. Two-way fixed effects anova and interactions
NWK readings for intro factorial designs
one-way anova NWK 16.1-16.9 ver 4; 16.1-16.6 ver5
post hoc pairwise comparisons NWK 17.4-17.5 ver4 and ver5; 
factorial designs: two-way fixed effects NWK 19.1-19.6, 20.2,20.3 ver4; 19.1-19.6, 19.9 ver5
B. More Factorial Designs
1. Random and mixed anova models (multiple comparisons, variance component estimation)
2. Unbalanced designs
3. k-way classifications
4. Design--Sample size and power
5. Randomized block designs (including Latin Squares)
NWK readings for more factorial designs
mixed and random 2-way NWK 24.2-24.4 ver4; 25.2-25.4 ver5
one observation per cell NWK 21.1-21.2 ver4 and ver5
Unbalanced two-way designs NWK 22.1, 22.2, 22.6 24.6 ver 4; 23.1, 23.2, 23.6 23.6 ver5; 
three-way factorial designs NWK 23.1-23.6, 24.5 ver4; 24.1-24.5, 25.6 ver5 
planned (orthogonal) comparisons NWK 17.3 ver4 and ver5  
design and sample size NWK 26.1-26.5 ver4; 16.10,16.11, 19.11, 24.7 ver5
randomized block designs NWK 27.1-27.7, 30.1-30.2 ver4; 21.1-21.9, ver5 
C. Nested and Repeated Measures Experimental Designs
1. Nested designs
2. Repeated measures designs
NWK readings for nested and repeated measures designs
nested and crossed-nested NWK 28.1-28.5, 28.9 ver4; 26.1-26.5, 26.9 ver5
repeated measures designs NWK 29.1-29.4 ver4; 27.1-27.4 ver5

II. Analysis of Association: Correlation and Regression

Correlation and Straight-line regression

A. Basic Regression Models
1. Multiple regression
2. Polynomial regression
3. Model violations and transformations
Note: readings for introductory regression lectures Part A
Review: Straight-line regression  NWK Ch 1-4 ver4,5
Multiple Linear Regression
  Basic fit: Inference for params & fit  Ch.6 ver4,5
  R-sq, adj R-sq pp230-1  ver 4; 226-7 ver 5
  Adjusted Variable Intepretation (partial regr) sec 9.1 ver 4; sec 10.1 ver5 (added-variable plots) 
  Testing composite Hypoth sec 7.1-7.3 ver4,5
  partial part correl sec 7.4 ver4,5
  standardized coeff sec 7.5  ver4,5
  polynomial regr sec 7.7 ver4; sec 8.1 ver5
  Inference for correlations sec 15.4 640-643 ver 4; sec 2.11 ver5
     heteroskedascity sec 10.1 ver 4; 11,1 ver 5; autocorrelation ch12.1-12.4 ver4,5;
     multicollinearity sec 7.6 ver4,5, VIF sec 9.5, 10.2 ver 4; sec 10.5 ver5
     outliers, resduals sec 9.2 ver4; sec10.2 ver5
B. Regression Models with Categorical Variables
1. Reformulation of anova models
2. Analysis of covariance & alternatives
Note: readings for regression lectures Part B: categorical predictor vars, 
   Qualitative predictors: NWK Ch 11 ver 4, Chap 8 ver5 ; 
   Ancova (via anova models)NWK Ch 25 ver4, Ch 22 ver 5
    Qualitative predictors:
     0,1 dummy vars, reg params sec 11.1 p456- ver4; 8.3 p.313- ver5
     non-parallel regressions  sec 11.2 ver4, sec8.4,8.5 ver5
     regr approach to ancova, more than 2 groups  sec 11.3 ver4 , sec8.6 ver5
     anova one-way sec 16.11, 2-way sec 19.7 p.832 ver 4
                   sec 16.8  ver5      
      reduction of error var sec 25.1 ver4, 22.1 ver5
      single factor sec 25.2, crackers ex sec25.3  ver 4, 22.2,3 ver5
C. Building Regression Models
1. Variable Selection and Model Construction:
          Statistical algorithims, stepwise regression, best subsets
          Composites and variable reduction (including principal components)
2. Model building by "theory", Intro Path Analysis and LISREL (see ed260 page, Rogosa "casual models")
3. Regression models with hierarchical data
Note: readings for regression lectures Part C: Model Building, 
  stepwise, best subsets, "automatic" NWK 8.1-8.5 ver 4, 9.1-9.5 ver 5
  cross-validation NWK 10.5-10.7 ver 4, 9.6 ver 5
  advanced topics: path analysis, hierarchical data see ed260 page

III. Analysis of Categorical Data

A.    Proportion and Count Outcomes:
        Intro and Review:  Bernoulli, Binomial, Multinomial, and Poisson distributions; inferences for proportion and count data;  
       Univariate Categorical Data; Logit and odds transformations;
       Generalized Linear Models: Logistic and Poisson Regression
         Readings for IIIA
         NWK Ch.14, ver4, ver5 Logistic regression, Poisson Regression
         Agresti Ch.1 (proportions and counts); 4, 5, 8 (logistic, poisson regression); 10 (history)

B.    Statistical Modelling, Estimation, and Inference for Multivariate Categorical Data
         Review: Basic contingency Tables
         Odds-ratios, conditional and marginal independence, Simpsons Paradox,
         Cochran-Mantel-Haenszel for metanalysis,
          Log-linear models for Multi-way Contingency Tables,
         Associations among ordinal variables
         Agresti Ch. 2, 3, 6, 7, 9.

Additional Readings

Bringing Evidence-Driven Progress To Education:
main report November 2002           US DOE press release       December 2003 confab, "what works"
Rogosa, D. R. (1980).   Comparing nonparallel regression lines.
  Psychological Bulletin, 88, 307-321.

Rogosa, D. R. (1987).  Casual models do not support scientific
   conclusions: A comment in support of Freedman.
   Journal of Educational Statistics, 12, 185-195.

Guest books