As far as I can judge, you are doing nothing wrong.
The easiest example of what you may have in mind is when you ask for summary statistics from several variables. It is then entirely possible to look at each variable separately and use all available values. But the results are then not strictly comparable if they are based on sets of observations with offlap.
Similarly, that can be done for pairs of variables from which you want lots of correlations.
However, such a procedure does not carry over to models of the kind you are using. If there are missings in an observation, then Stata can do nothing with that observation. One way to see this is to think of the underlying matrix operations: all elements in all matrices must be present.
As Maarten and I both pointed out, the only way forward that differs from Stata's behaviour here is to use imputation for your missings.
That's my understanding of your problem.
Nick
[email protected]
Jessica [iso-8859-1] �lschl�ger
Dear Nick,
I don't want Stata to "try harder".
I just wanted to make sure that I don't do anything wrong.
My "doctor father" (supervisor) mentioned to me that in some statistics
packages
there is a way to use every observation available (but he doesn't use Stata),
whereas I think Stata makes some kind of balanced panel out of my dataset when
there are missing observations for one country i at time t, i.e. doesn't use
any
of the observations available for i at time t because of the missings.
As I am still a beginner in Stata, I may as well be mistaken at this point.
But IF I'm doing something wrong, I really would appreciate somebody telling
me.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/