st: RE: Stata date format %tw: how long is the weekend in a 9-day week?

[Nick Cox <n.j.cox@durham.ac.uk>]

Please see 

SJ-10-4 dm0052  . . . . . . . . . . . . . . . . Stata tip 68: Week assumptions
        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N. J. Cox
        Q4/10   SJ 10(4):682--685                                (no commands)
        tip on Stata's solution for weeks and on how to set up
        your own alternatives given different definitions of the
        week

For Stata, week 1 of any year starts on 1 January, regardless of what day it is.

Week 52 may have (indeed will have) more than 7 days. 

These are unlikely to fit with many needs, and the Tip above is, in the absence of any apparent competition, all that is written on the subject. I agree, however, with your implication that this should all be better documented in the manual. The Stata 12 manuals do reference the Tip above. 

If we go back to basics, the facts are that weeks do not nest inside years or months and that practical definitions of week differ a lot, so Stata is agnostic on the point. 

Nick 
n.j.cox@durham.ac.uk 


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Sergiy Radyakin
Sent: 29 July 2011 16:50
To: statalist@hsphsun2.harvard.edu
Subject: st: Stata date format %tw: how long is the weekend in a 9-day week?

Dear All,

I have a question regarding the Stata's %tw format. The manual (help
dates_and_times) says in this format the numeric values are
interpreted as:

"a %tw value records the number of weeks from the first week of 1960"

I have the following questions:

1) The year 1960 started on a Friday. Why is Wednesday the 6th of Jan
1960 is still considered the first week of 1960?

See the following code, which shows the day-to-week assignment in the
month of January 1960:
forval i=0/30 {
  display wofd(`i')
}

Reformulating the question is the week reported in the format like 1975w9
  A) the calendar week?
  B) the number of the 7-day-long group since the beginning of the
year where a particular date belongs to?
  C) something else?


2) The second question comes as we depart from 1960 and move on to the
current dates. For an example take a value like 1975. The date that is
1975 weeks away from the Jan 1, 1960 falls somewhere in early November
1997, while Stata reports it as the 52 week of the said year. I can
substantiate the question with the example in Excel:

Format a cell with the default date format. Enter the value 21916,
which corresponds to Jan 1, 1960 (in the US notation 1/1/1960)
Format another cell with the default date format. Enter a formula
=A1+7*1975 - which will add 1975 weeks (of 7 days each) to the first
date. The result (in the US notation) should be 11/7/1997.


3) My final question is related to the length of the year in weeks
assumed by Stata: I have never seen Stata reporting the value YYYYw53
in any of the years, which is rather strange, since any year will
contain dates beyond the 52 week. The investigation with the following
code led to very interesting results:

forval i=0/365 {
  display `i' " " wofd(`i')
}

Here 0-365 are the 366 days of the good old year 1960 which was a leap
year. Interestingly, according to Stata the year ended in a long week
consisting of 9 days!!!
357 51
358 51
359 51
360 51
361 51
362 51
363 51
364 51
365 51
(I just hope there was a very-long weekend :)

Jan 1, 1961 (day 366) is then assigned week 52 and it lasts for 7 days as usual.

The manual declares the valid range for week number 1-52, which agrees
with the results above, but doesn't give any insight into the fact
which week of the year will have 8 or 9 days.

I don't think there is a bug in Stata with regards to the handling of
dates (though the last issue illustrates an apparent unorthodox
approach), but rather I seek a more elaborate explanation of how Stata
comes to the (formatted) values like "1997w52" for value 1975 and the
format %tw. The manual for date-time handling is rather cumbersome and
is more elaborate about atomic time and leap seconds rather then 9-day
weeks! Perhaps it can be improved by an easy "getting started" page
that would deal with only one (recommended) format for date and the
rest is put into a different reference-style topic.

The version of Stata does not play a role here and the quoted results
are reproducible in Stata v9-11 (Windows).

Thank you,
    Sergiy Radyakin
*
*   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/

*
*   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/