st: correcting errors in panel TSCS survey

["DeMichele, Matthew" <mdemichele@rti.org>]

I'm working with a survey collection across US states (n = 50) conducted
from 1980-2010 (t=21) in which respondents were asked about the
population of their hospital. Populations are recorded for the end of
each year (December 31) and for the beginning of the following year
(January 1). Although there is only a 1 day difference, sometimes there
are large differences between these numbers. These differences could be
due to new policies, shifting people from different wards, and other
administrative changes. This creates a situation in comparing changes in
populations across years becomes problematic. I am working to correct
for these discrepancies, but am a bit unsure of exactly how to do this.
Also, these sort of administrative changes will cause similar
discrepancies for future waves. I have notes from respondents that
provide the reason(s) for large discrepancies in end of year and
beginning of year hospital populations, and hoped to provide a suggested
weighting scheme for the different types of discrepancies. I may be far
off base here, but is this is a matter of smoothing or forecasting? Or,
do stata listers have different ideas? 

Thank you in advance for your help.

Matthew 



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