Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Multiple imputation in panel data when subjects die

From   Morten Hesse <>
Subject   Re: st: Multiple imputation in panel data when subjects die
Date   Fri, 10 Sep 2010 12:58:39 +0200

variable for being dead in a given year (which will be one for each year). After you have re-reshaped to long, replace imputed values with missing.
Hope this helps.
Den 10-09-2010 12:45, skrev:
Dear Statalist

I have a panel data set with some missing values which I would like to impute using Stata's mi command. However, over time, subjects in my panel die.

An example of the type of pattern I observe is:

Subject 1: M M M O O O D D D D D D

Subject 2: O O O O O M O O O O O D
Where M is 'missing', O is 'observed' and D is 'dead'.
In the exchange between Yulia Marchenko and Jibonayan Raychaudhuri (, Paul Allison's (2001, page 74) approach to dealing with missing values in longitudinal data is outlined, namely, if the data set is in "long" form, reshape it to "wide" form so that there is one record for each subject (with distinct variables for measurements on the imputation variable at different points in time) and then perform the imputation, before reshaping back to "long" form.

Because, over time, my subjects die, I have some missing values ("."s) which need to be imputed, because they are "true missing values" (the "M"s above), and also missing values which should not be imputed (the "D"s).

My proposed solution is to replace all missing values owing to the subject having died (the "D"s) with another Stata coding for a missing value (e.g. ".a"), so that only the true missing values (the remaining "."s) are imputed by mi.

Taking this approach, and using the reshaping approach suggested by Allison and Yulia outlined above, Stata successfully imputes missing values for the "."s and not the ".a"s, which I think is great.

My question is this: does my approach make sense, that is, does it represent a "principled" approach to mi for panel data in the presence of deaths, in the spirit of Allison (2001) and Yulia?

With thanks, in advance, for any help anyone can give me.
Martin Forster

REF: Allison, P. 2001, Missing Data. Sage University Papers Series on Quantitative Applications in the Social Sciences.

*   For searches and help try:
*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index