st: Oddities with weekly data

["Michael S. Hanson" <mshanson@mac.com>]

	I have a student who is working with weekly data on equity returns.  
When he attempts to -tsset- his data based on his "time" variable, he 
gets the error:

repeated time values in sample
r(451);

	Stata seems to have trouble with two cases, as illustrated in the 
listing below.  First, most of his observations are Mondays, but in 
some cases the recorded data is on a Tuesday. (Presumably the market 
was closed;  the dates do appear to coincide with US federal holidays, 
but I'm puzzled why there aren't many more of them in his sample.)  
Stata is apparently treating Monday as the end of the week, and so 
Tuesdays are assigned to the following week -- meaning both a duplicate 
week and a missing week.  I'd prefer finding a way to convince Stata 
that a week is defined by any value in a Monday-Friday interval, but 
I'd settle for a quick way to recode all the Tuesdays as Mondays.

	Second, in 1990, 1996 and 2001, the last Monday of the year is the 
53rd.  Stata (understandably) does not like this structure of the data. 
 I'm at a loss of how to deal with this issue, however, as all the 
years in the sample (appear to) have 52 weeks -- or more -- so he 
cannot just shift a week from one year to another.  (The data range 
from Feb 1998 to March 2006.)  Surely some basic accounting has gone 
awry here?  A reference was made to this problem earlier on Statalist 
(see <http://www.stata.com/statalist/archive/2003-05/msg00129.html>), 
but I cannot think of any concrete ways to address it.

	Suggestions on ways to approach both aspects of this problem from 
within Stata would be appreciated;  I'd rather not have to recode the 
data by hand if that is avoidable.  I've asked the student to work with 
time series operators and estimators on these data ;  hence the need to 
-tsset- them as weekly.  Thanks!


. list date time dow if time == time[_n-1] | time == time[_n+1], sep(2)

     +----------------------------------+
     |       date      time   dowretu~s |
     |----------------------------------|
151. | 12/24/1990   1990w52   -.0016911 |
152. | 12/31/1990   1990w52   -.0243001 |
     |----------------------------------|
159. |  2/19/1991    1991w8   -.0155532 |
160. |  2/25/1991    1991w8    .0070837 |
     |----------------------------------|
173. |  5/28/1991   1991w22    .0382414 |
174. |   6/3/1991   1991w22   -.0169085 |
     |----------------------------------|
187. |   9/3/1991   1991w36   -.0105596 |
188. |   9/9/1991   1991w36   -.0086506 |
     |----------------------------------|
464. | 12/23/1996   1996w52      .01173 |
465. | 12/30/1996   1996w52    -.002567 |
     |----------------------------------|
724. | 12/24/2001   2001w52    .0100782 |
725. | 12/31/2001   2001w52    .0120364 |
     |----------------------------------|
728. |  1/22/2002    2002w4     .006958 |
729. |  1/28/2002    2002w4     .006804 |
     |----------------------------------|
732. |  2/19/2002    2002w8    .0065532 |
733. |  2/25/2002    2002w8    .0394121 |
     |----------------------------------|
746. |  5/28/2002   2002w22   -.0178751 |
747. |   6/3/2002   2002w22   -.0343955 |
     |----------------------------------|
760. |   9/3/2002   2002w36   -.0276542 |
761. |   9/9/2002   2002w36   -.0136813 |
     +----------------------------------+

                                        -- 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/