How to generate a variable using dates in the IF function

[Danielle Berg]

Dear Stata users,

I'm having trouble generating a new variable in Stata. This is the current command I run:

gen regulation=.

replace regulation=0 if DateAnnounced<=9/2/1992 & TargetNation=="AS"

However, Stata mentions that 0 real changes are made. However, I do know that in my dataset are observations present with the TargetNation being "AS" (Austria) and are announced before september 2nd, 1992 (I found out by running a command where I used my Year variable instead of my DateAnnounced variable. If I look at the variable DateAnnounced, I see in the top box that Stata sees the variable as a date, so I figured Stata would also know if I want observations on that date and before. However, this does not seem to work. I need dates instead of years as I need specific days on which I want to run my regression. Does someone know why Stata doesn't seem to get dates if I use the <= and then put in the desired date? (see screenshot for how the variable looks)

Thank you in advance!

Kind regards,
Danielle

[Nick Cox]

if here is a qualifier, not a function.

The difficulty is that Stata does not know that you think 9/2/1992 is a date. It thinks you're asking for divisions. You can check how it is thinking directly:

. display 9/2/1992
.00225904

. display %td  9/2/1992
01jan1960

So the syntax is legal and yields a number that in date terms is 1 January 1960, which clearly is not what you want.

The cardinal rules for daily dates are that

1. daily dates are held as integers with 0 = 1 January 1960

2. the display format is something completely different.

What you want can be got in various ways, the easiest to my mind being

if DateAnnounced<= mdy(9, 2, 1992) 

with the rest of your syntax looking to be legal so far as I can see.

As a matter of taste I go for 0, 1 indicators as explained at length in

https://www.stata.com/support/faqs/d...rue-and-false/

https://www.stata-journal.com/articl...article=dm0087

https://www.stata-journal.com/articl...article=dm0099

You could get such a variable directly with

gen whatever =  DateAnnounced<=mdy(9,2,1992) & TargetNation=="AS"

[Danielle Berg]

if here is a qualifier, not a function.

The difficulty is that Stata does not know that you think 9/2/1992 is a date. It thinks you're asking for divisions. You can check how it is thinking directly:

. display 9/2/1992
.00225904

. display %td 9/2/1992
01jan1960

So the syntax is legal and yields a number that in date terms is 1 January 1960, which clearly is not what you want.

The cardinal rules for daily dates are that

1. daily dates are held as integers with 0 = 1 January 1960

2. the display format is something completely different.

What you want can be got in various ways, the easiest to my mind being

if DateAnnounced<= mdy(9, 2, 1992) 

with the rest of your syntax looking to be legal so far as I can see.

As a matter of taste I go for 0, 1 indicators as explained at length in

https://www.stata.com/support/faqs/d...rue-and-false/

https://www.stata-journal.com/articl...article=dm0087

https://www.stata-journal.com/articl...article=dm0099

You could get such a variable directly with

gen whatever = DateAnnounced<=mdy(9,2,1992) & TargetNation=="AS"

Thank you so much for the fast reply! It worked