Re: st: lincom & svy commands

[Steven Samuels <sjhsamuels@earthlink.net>]

Stas has provided a detailed discussion, and  Maarten has already  
shown you the formula that -lincom- uses. If you do not understand  
the relationship between -lincom- (a t-test) and the regression  
approaches that Stas presented, consult the text for any graduate  
statistics course which includes regression.  I will answer your  
second question.

Alternative ii will be more precise in your example.  Analyses with  
clustered data (from surveys or not) should never ignore the clustering.

When you examine difference within the same cluster, each cluster  
serves "as its own control".  Here is a simple example with 4  
clusters. Assume that the number of observations within cluster is  
equal.

Cluster	  Before-Mean	After-Mean   difference
1           5              4          1
2          10              8          2
3          15              11         4
4          20              16         4

The differences  are much less variable (range = 4) then the original  
observations (ranges = 15, 12).

What are the consequences? If you apply your option i, then:  n= 8,  
and the before-after difference is 2.75, with a SE = 4.10  and the  
two-sample t-test has p = .52

If you treat the data as clustered (option ii), the before/after  
difference is still 2.75 with n=4 clusters, but, the se = 0.75 and  
the one-sample paired  t-test has p= 0.03


Now, this is just an ordinary paired t-test example utilizing the  
cluster means as responses.  Stata's survey programs account for  
unequal sample-sizes within cluster and occasion. You ask for what - 
lincom- (or the regression equivalents) "actually" do.   The  
equations are a bit complicated and depend on the choice of variance  
estimate (linearization, jackknife, bootstrap, balanced-repeated- 
replications).  You can view generalized versions of the equations in  
the Stata survey manual or in any book on survey sampling.  Take a  
look, for example, at Sharon Lohr, Sampling: Design and Analysis.  
Duxbury, 1999, Chapter 9.  However, the logic with before/after data  
is the same.




-Steve



On Oct 23, 2008, at 11:18 AM, Bell, Jacqueline S. wrote:

> Hi
>
> This is a follow-on to a previous message I sent this month asking  
> about how lincom calculates standard errors when clustering is  
> present.
>
> Can anyone advise me on what lincom actually does when estimating  
> differences in parameters from svy:mean or svy:prop?
>
> I have before/after data which is not paired at an individual  
> level, but has a cluster structure.  In these circumstances it is  
> not obvious how lincom goes about estimating the before/after  
> difference.
> The two alternatives suggested to me are:
> i) it estimates before and after separately for the whole  
> population, then estimates the difference
> ii)it estimates the difference in each cluster, and then the  
> overall difference.
>
> In the data there are (in most cases) before and after data for  
> each cluster, but often quite severe imbalances in samples before/ 
> after within cluster.
>
>
> Thanks for any help, Jacqui
>
>
>
> The University of Aberdeen is a charity registered in Scotland, No  
> SC013683.
>
> *
> *   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/

*
*   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/