My approach (generating your own week variable, essentially, bypassing
the wofd() function) would solve the Mon/Tue problem and the week 52
problem, but assuming there are no missing weeks, your alternative of
using _n seems just as good or better. The oddity of weeks is that
Stata constrains there to be only 52 weeks in a year, and the week
numbering starts over on Jan 1 of each year, which means that week 52
in any year is either 8 or 9 days long. Having repeated time values
in a weekly time series is nearly inevitable given these definitional
constraints.
clear
range t 12040 12060 21
gen twk = wofd(t)
gen d=t
format d %d
li, noo sep(7)
On 5/18/06, Michael S. Hanson <[email protected]> wrote:
As an alternative, I could just -tsset- with a simple time trend:
//
sort t
gen trend = _n
tsset trend
//
...but that has the disadvantage that the time variable doesn't have
any natural units; thus one cannot use the -tin()- functions for
selecting subsamples, dates don't show up correct on the t-axis of
graphs, etc.
So I'm still a bit stuck -- something apparently is still "odd" with
these data -- but I do appreciate your help. As I mentioned above, any
additional thoughts would be most welcome. Thanks again.
-- Mike
*
* 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/
