/PM-MI

code for 'Planned Missingness with Multiple Imputation: an Application to Election Day Surveys'

Primary LanguageStata

Code for data processing, multiple imputation and analysis for a method that couples Planned Missingness with Multiple Imputation on Exit Polls

The method was originally introduced in:

Bautista, René, Marco A. Morales, Francisco Abundis and Mario Callegaro. 2008. Exit polls as valuable tools to understand voting behavior: Using an advanced design in Mexico - Measurement Error. In Elections and Exit polling, ed. Fritz Scheuren and Wendy Alvey. New York, NY: John Wiley & Sons

and further upacked/developed in the companion paper 'Planned Missingness with Multiple Imputation: an Application to Election Day Surveys'

Each file on the repo performs the particular tasks described below:

  • exitpoll06.do : performs initial data manipulations to generate file in preparation for MI (Stata do-file)
  • splitmex06_MI.R : perfoms multiple imputation on the full data file using Amelia II (R file)
  • MI_reshape.do : performs computations and data reshaping to long format on each multiply imputed file (Stata do-file)
  • exitpoll06_pres.do : performs vote choice analysis using multinomial probit model on each of the MI datasets and combines results using 'Rubin rules' (Stata do-file)
  • exitpoll06_sim_pres.do : generates simulations from estimated models to compute diverse first differences of interest (Stata do-file)
  • coef_graphs_110717.do : graphs first differences on a single file appearing as Figure 3 on the paper (Stata do-file)