st: Partitioning Variance for Single Variables in Complex Survey Data

[Lloyd Dumont <lloyddumont@yahoo.com>]

Hello.  I am using a panel which I can analyze either as complex survey data or as cross-sectional time series.  Each person is observed twice-�rongly balanced.  I have run the regressions (OLS and probits) approaching the data as survey (rather than xt) data, so that is probably the better way for me to approach the following problem.

For binary variables, I would like to be able to say…
Of all those that had X=1 at time 2, almost all of them already had it at time 1.  In other words, most of the variance in X is not within-, but rather between-people.

For continuous variables, I would like to be able to say…
There is a very high correspondence between a person’s value for X in year 2 and the same person’s value for X in year 1.  Again, I want to be able to say that most of the variance in X is not within-, but rather between-people.

If I were to rely on xt commands, I have a sense of how I would handle the binaries.  I could run xtsum and xttrans.  However, even those do not allow for percentage breakdowns in within- and between- variance, I don’t think.  Even if they did, this leaves open the issue of the continuous variables and does not respect the complex sampling (i.e., survey design) of the data.

Thank you for your suggestions.  Lloyd Dumont



      


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/