st: Interactions between tsset and sort

[Gordon Hughes <G.A.Hughes@ed.ac.uk>]

I would like to raise a point for general discussion and perhaps 
consideration by Stata Corp.  It concerns the way in which --tsset-- 
sorts a panel dataset.  As far as I understand, using --tsset panel 
time-- is equivalent to --sort panel time-- with respect to the sort 
order of the data.  A later sort such as --tsset time panel-- cancels 
the tsset status of the data.

This is the right thing to do in 95% of cases because it ensures that 
time series operators work correctly.  However, there is a class of 
circumstances in which this behaviour has unwanted consequences.  I 
am dealing with spatial panel data for which the natural way to stack 
panel data corresponds to sorting by time, then panel so that spatial 
correlations between panel (spatial) units can be dealt with more 
easily.  At the same time for user-written routines it is still 
necessary to have the data --tsset-- in order to identify the panel 
and time variables easily.

To get round the problem I use --tsset, noquery-- to identify the 
panel and time variables, then -sort time panel-- and then finally 
--tsset panel time-- at the end of the ado file.  This is ok if 
everything works but it leaves a mess behind if the routine exits 
with an error.  It would be much easier if --tsset-- had an option - 
e.g. [nosort] - which leaves the sort order of the data untouched, 
but also tells Stata that the data is not time series data so that 
time series operators cannot be used.

Gordon Hughes
g.a.hughes@ed.ac.uk

*
*   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/