st: time smoothing in panel data

[Rufus Peabody <rufus.peabody@gmail.com>]

I'm attempting to create a model to predict performance.  I have about  
500 different groups, and have as many as 200 and as few as 1  
observation for each group.  These occur over an 8 year period and  
there are gaps in days between them.  For example, one group might  
have an observation July 1, then July 6; for another group, it could  
be July 2, then July 9.   What  I want to do is determine how much I  
should weigh each previous observation to predict performance for a  
future observation.  I would have thought -tssmooth nl-  might work,  
but it did this by group.  I want to control for group and create a  
decay function that I can use to predict future performance for all  
groups.  Any help would be greatly appreciated!

Thanks,
Rufus
*
*   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/