... University¹

© Copyright 1995-2006, All rights reserved. (Institute for Quantitative Social Science, Harvard University, 1737 Cambridge St., Cambridge MA 02138; http://GKing.Harvard.Edu, King@Harvard.Edu, (617) 495-2027.) This document can be read on-line in hypertext format from http://GKing.Harvard.Edu/ei/ei.html. The program implements the procedures described in A Solution to the Ecological Inference Problem: Reconstructing Individual Behavior from Aggregate Data (Princeton University Press, 1997). tex2html_wrap_inline$EI$ was written in the Gauss Programming Language, ©Copyright Aptech Systems, Inc., and uses Gauss' Constrained Maximum Likelihood Module written by Ronald J. Schoenberg. For making available public domain Gauss code, I am grateful to David Baird (for an inverse cumulative normal procedure), Martin van der Ende (for an accurate cumulative bivariate normal procedure), and Simon Jackman (for a loess procedure). For research assistance, my thanks goes to Kosuke Imai and Eric Dickson, and for research support, my thanks goes to the National Science Foundation (IIS-9874747), the National Institutes of Aging (P01 AG17625-01), and the World Health Organization. EI is copyrighted, but you may copy and distribute this program provided that no charge is made and the copy is identical to the original. To request an exception, please contact me.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...n).²

For example, to read in the data from an ascii text file, you can include these two commands after the library command: clear t,x,n; and loadvars sample.asc t x n. A detailed example appears on page .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... same.³

Also stored in the output data buffer are the data input to this procedure, and all global options chosen during the analysis. As such, this data structure is very convenient if you are in the habit of following the replication standard by making publically available all the data and information necessary to replicate your published analyses. In fact, the output data buffer is an automatically created and self-documented ``replication data set.'' To follow the replication standard, you only need to provide this ${\mathfrak EI}$ data buffer to the ICPSR or some other public archive; others will then have access to the data and information necessary to replicate your results (of course, they won't have any other information unless you decide to provide it). To replicate an ${\mathfrak EI}$ analysis, you only need to load in the data buffer, loadm dbuf;, and call this procedure: dbufNew=eirepl(dbuf);. See Gary King, ``Replication, Replication,'' PS: Political Science and Politics, with comments from nineteen authors and a response, Vol. XXVIII, No. 3 (September, 1995): 443-499.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.