Generalized Additive Selection Models for the Analysis of Non-ignorable Data Software

Download archive with all files.

BIAS, VARIANCE, MSE SIMULATIONS

These are the files containing all the source code needed to run the simulations in Scharfstein and Irizarry's "Generalized Additive Selection Models for the Analysis of Non-ignorable Missing Data" The functions performing the estimation are in gam-fit.S, doubly-robust.S, and orthogonal.S

simulation.S The simulation can be started by simply sourcing this file in Splus 3.4
functions.S Defines some Splus function used throughout
constants.S This file contains the definitions of most constants, such as number of data points, effect sizes, etc...
model-generation.S generates the models determined by the information in constants.S
data-generation.S creates the data, Y, X and R using models defined above
gam-fit.S performs the gam-type procedure to estimate gamma
doubly-robust.S computes the double robust estimates
orthogonal.S computes the orthogonal estimate you can then run
show.S shows the results of simulation you can use
test.S gives some summary stats for the data made with model generation and data-generation


BOOTSTRAP, MODEL SELECTION SIMULATIONS
bootstrap.S you can run the bootstrap simulations by simply sourcing in into Splus 3.4
bootstrap-gam-fit.S for these simulation the gam fit is a bit different because we need to predict