show only first day of quarter

[Eva Smith]

please help me with this table

            01jan2008
23.17    02jan2008
22.49    03jan2008
23.94    04jan2008

25.71    28mar2008
25.61    31mar2008
22.68    01apr2008
23.43    02apr2008
23.21    03apr2008
22.45    04apr2008
22.42    07apr2008
22.36    08apr2008
22.81    09apr2008
21.98    10apr2008
23.46    11apr2008
23.82    14apr2008
22.78    15apr2008

21.58    19jun2008
22.87    20jun2008
22.64    23jun2008
22.42    24jun2008
21.14    25jun2008
23.93    26jun2008
23.44    27jun2008
23.95    30jun2008
23.65    01jul2008
25.92    02jul2008
24.78    03jul2008
            04jul2008
25.78    07jul2008
23.15    08jul2008
25.23    09jul2008
25.59    10jul2008
27.49    11jul2008
28.48    14jul2008
28.54    15jul2008

33.85    22sep2008
35.72    23sep2008
35.19    24sep2008
32.82    25sep2008
34.74    26sep2008
46.72    29sep2008
39.39    30sep2008
39.81    01oct2008
45.26    02oct2008
45.14    03oct2008
52.05    06oct2008
53.68    07oct2008
57.53    08oct2008
63.92    09oct2008
69.95    10oct2008
54.99    13oct2008
55.13    14oct2008

... 
and other years

in this table i should drop all, except only first day of quarter, and if first day is missing it should be replaced by second date

23.17	01jan2008
22.68	01apr2008
23.65	01jul2008
39.81	01oct2008
etc	01jan2009

[Clyde Schechter]

So, let's call the first variable v1 and the other variable date. I'm going to assume that this is a genuine numeric Stata internal date variable, not a string variable that humans read as dates, and not some other kind of number with value labels attached. You want to keep only those dates that represent the first date in each quarter

gen quarter = qofd(date)
format quarter %tq

by quarter (date), sort: keep if _n == 1

This will leave you with only the earliest date in the quarter your data set contains.

The way you posed your question leads me to wonder, though, whether you want to go farther and retain that date only if it is the first or second calendar date in the quarter (so that a quarter would go completely unrepresented if, say, the first date from the quarter in the data set is the third date in the quarter or later.) If that's what you want, then follow-up the above with the following:

drop if date-dofq(quarter) >= 2

See online help for date and time functions, and the manual section for (a lot) more detail. The key here is the use of the qofd() function (and perhaps the dofq() function).

[Eva Smith]

...

01jan2008 23.17 02jan2008
small problem is in first jan, it is not working day and at all years obs missing. How replace obs value from 2nd jan to first

and better if code can find all missing obs and replace by second obs

[Nick Cox]

Clyde's post answered that already.

[Eva Smith]

v1   date
               01jan2008
22.68    01apr2008
23.65    01jul2008
39.81    01oct2008
            01jan2009
42.28    01apr2009
26.22    01jul2009
28.27    01oct2009
             01jan2010
17.47    01apr2010
32.86    01jul2010
22.5    01oct2010
17.4    01apr2011
15.87    01jul2011
             02jan2012
15.64    02apr2012
16.8    02jul2012
16.32    01oct2012
              01jan2013
13.58    01apr2013
16.37    01jul2013
15.54    01oct2013
            01jan2014
13.1    01apr2014
11.15    01jul2014
16.71    01oct2014
            01jan2015
15.11    01apr2015
16.09    01jul2015
22.55    01oct2015
            01jan2016

i dont understand. i tried

 
 drop if date-dofq(quarter) >= 2

it deleted all obs after 2016q1, missing values did not replaced

v1          date    quarter
16.6    02oct2013    2013q4
           01jan2014    2014q1
14.23    02jan2014    2014q1
13.1    01apr2014    2014q2
13.09    02apr2014    2014q2
11.15    01jul2014    2014q3
10.82    02jul2014    2014q3
16.71    01oct2014    2014q4
16.16    02oct2014    2014q4
             01jan2015    2015q1
17.79    02jan2015    2015q1
15.11    01apr2015    2015q2
14.67    02apr2015    2015q2
16.09    01jul2015    2015q3
16.79    02jul2015    2015q3
22.55    01oct2015    2015q4
20.94    02oct2015    2015q4
             01jan2016    2016q1

[Clyde Schechter]

I think there has been a misunderstanding. When Eva referred to "missing," I interpreted that to mean that there was no observation with that date. It now appears she means that v1 has a missing value for that observation. And when that happens, she wants to use the first non-missing value for the quarter. So I think what she wants is this:

drop if missing(v1)
gen quarter = qofd(date)
format quarter %tq

by quarter (date), sort: keep if _n == 1

This will leave her with a reduced data set containing one observation per quarter. That observation will be the earliest date in the quarter that has a non-missing value for v1.

It is not clear to me whether she also wants to change the value of date to be the first day of that quarter, even though the value of v1 actually comes from a later date. If she does want to do that, she can follow the above code with:

gen new_date = dofq(quarter)
format new_date %td