Data
====

1. congress.data.R
  - N            : number of observations
  - incumbency_88: indicator of whether the current occupant of the congressional seat
		 is running for reelection
                 +1: district is currently occupied by Democrat who is running for 
		     reelection
                  0: election is open (neither of the two candidates is currently 
                     occupying the seat)
                 -1: district is currently occupied by Republican who is running for 
		     reelection
  - vote_86      : democratic share of the two-party vote in 86 election
  - vote_88      : democratic share of the two-party vote in 88 election

2. earnings.data.R
  - N       : number of observations
  - earnings: earnings in dollars
  - height  : height in inches
  - sex     : 1: male, 2: female

3. earnings1.data.R
  - N       : number of observations
  - earn_pos: is earnings positive? 1: Yes, 0: No
  - height  : height in inches
  - male    : is male? 1: Yes, 0: No

4. earnings2.data.R
  - N       : number of observations
  - earnings: earnings in dollars
  - height  : height in inches
  - sex     : 1: male, 2: female

5. wells.data.R
  - N      : number of observations
  - arsenic: level of arsenic of respondent's well
  - assoc  : any household members active in community organizations? 1: Yes, 0: No
  - dist   : distance (in meters) to closest known safe well
  - educ   : education level of head of household
  - switc  : household switched to new well? 1: Yes, 0: No

Models
======

1. One predictor
  wells.stan: glm(switc ~ dist, family=binomial(link="logit"))

2. Multiple predictors with no interaction
  congress.stan : lm(vote_88 ~ vote_86 + incumbency_88)
  earnings1.stan: glm(earn_pos ~ height + male, family=binomial(link="logit"))

3. Log transformations
  earnings_interaction.stan: lm(log(earnings) ~ height + male + height:male)
  earnings2.stan           : lm(log(earnings) ~ height + male)

