Data
====

1. electric.data.R
  - N         : number of observations
  - n_grade   : number of grades
  - n_pair    : number of pairs
  - grade     : grade levels
  - grade_pair: grade and pair interaction
  - pair      : pair number
  - pre_test  : pre test scores
  - treatment : received treatment? 1: Yes, 0: No
  - y         : post test scores

2. sesame.data.R
  - J      : number of site and setting combinations
  - N      : number of observations
  - siteset: combinations of site and setting
  - yt     : matrix of outcome y (post treatment score on test) and encouragement 
      	     T (whether child watched)
  - z      : pre treatment variable

Models
======

1. Linear model with one predictor
  electric_one_pred.stan: lm(post_test ~ treatment)

2. Linear model with multiple predictors and without interaction
  electric_multi_preds.stan: lm(post_test ~ treatment + pre_test)

3. Multilevel model with varying intercept
  electric.stan: lmer(y ~ treatment + (1 | pair))

4. Multilevel model with varying slope and intercept
  electric_1a.stan: lmer(y ~ 1 + (1 | pair) + (treatment | grade))
  electric_1b.stan: lmer(y ~ treatment + pre_test + (1 | pair))
  electric_1c.stan: lmer(y ~ 1 + (1 | pair) + (treatment + pre_test | grade))

5. Above models with Matt trick
  electric_1a_chr.stan: lmer(y ~ 1 + (1 | pair) + (treatment | grade))
  electric_1b_chr.stan: lmer(y ~ treatment + pre_test + (1 | pair))
  electric_1c_chr.stan: lmer(y ~ 1 + (1 | pair) + (treatment + pre_test | grade))
  electric_chr.stan   : lmer(y ~ treatment + (1 | pair))

6. Multilevel models with multivariate distribution
  sesame_street1.stan
  sesame_street2.stan
