Subtract 9 months from each of the 10K+ dates in a column

[Himanshu Singhai]

Hello, I've been trying to Subtract nine months from each of the 10K+ dates in a column and want to save it in another column. Could someone please help me with the code?

[Nick Cox]

For column read variable.

Are these monthly dates, daily dates, or something else?

If they are daily dates, what do you want to happen if 9 months before is 29 February in a non-leap year?

[Nick Cox]

There are more problems possible. Check if I have all cases:

31 Jan implies 31 April 9 months earlier

31 Mar implies 31 Jun ditto

31 Aug implies 31 Nov

30 Nov implies 30 Feb

29 Nov implies 29 Feb -- only in leap years

Here is one fudge demonstrated:

clear 
set obs 365 
gen ddate = mdy(12, 31, 2021) + _n 
format ddate %td 

gen guess = dofm(mofd(ddate) - 9) + day(ddate) - 1 
format guess %td 

bysort guess (ddate) : gen N = _N 

sort ddate guess 

list ddate guess if N > 1 

     +-----------------------+
     |     ddate       guess |
     |-----------------------|
 31. | 31jan2022   01may2021 |
 32. | 01feb2022   01may2021 |
 90. | 31mar2022   01jul2021 |
 91. | 01apr2022   01jul2021 |
243. | 31aug2022   01dec2021 |
     |-----------------------|
244. | 01sep2022   01dec2021 |
333. | 29nov2022   01mar2022 |
334. | 30nov2022   02mar2022 |
335. | 01dec2022   01mar2022 |
336. | 02dec2022   02mar2022 |
     +-----------------------+

[Nick Cox]

Another rule would be to subtract 274 days which is the best approximation to 9/12 of a year over a 4 year cycle:

. mata
: 4 * (273, 274, 275) / (3 * 365 + 366)
                 1             2             3
    +-------------------------------------------+
  1 |  .7474332649   .7501711157   .7529089665  |
    +-------------------------------------------+