Create time intervals given date (including time)

[Francois Durant]

Hi,

I would like to create 3 time intervals with a dummy variable called "period":
period = 1: 12:00:00 and below
period = 2: 12:00:00 to 14:00:00
period = 3 :14:00:00 and above

I have a variable "date" that contains the date and time, properly entered in Stata. Here is an example:
date
03feb2015 15:33:30

Here is the code (or pseudo-code) that I have so far:
* Break down by intervals:
gen period = 1 if date < 10:00:00
replace period = 2 if missing(period) & date <= 12:00:00
replace period = 3 if missing(period) & date <= 14:00:00
replace period = 4 if missing(period) & date > 14:00:00

Anyone can help? Thanks a lot!

[Nick Cox]

All you have to do is

1. read help dates and times

2. play with examples

3. work out what you need

1 and 2 give me in a few seconds

 
. di   hh(clock("24 Feb 2015 09:01:00", "DMY hms"))
9

from which what you need in your first example appears to be something like

 
gen hours = hh(mydatetime) 
gen mycat = cond(hours < 12, 1, cond(hours < 14, 2, 3))

I say "something like" as

a. What happens if the time is exactly 12:00:00? You say slightly different things. You may need other functions if getting it right to the nearest minute, second or millisecond really matters.

b. Your examples are different.

c. Such a variable would widely not be called a dummy variable.

but the details are up to you.

[Francois Durant]

Nick, thanks a lot for your response. That really helps!

[Francois Durant]

How would you modify the code if you wanted to create thin intervals thoughout the day? Say 15 minute intervals? Is there a more efficient way to achieve that? Thanks!

[Nick Cox]

I would use some variation on k * floor(datetime/k) where k is say 15 * 60 * 1000

[Francois Durant]

Would you be able to provide a bit more details on how to set k? Where is the 1000 from for instance? What would k look like with, say 90 minutes intervals? Thanks a lot!

[Nick Cox]

I didn't know the solution was so inscrutable, but the solution for 15 minutes was just making explicit that

1000 milliseconds = 1 second
60 seconds = 1 minute
etc.

otherwise someone might ask where 900000 appeared from.

The solution for 90 minutes could be just the solution for 15 minutes with 15 changed to 90 if you were satisfied with the results.

[Dr. Iqra Yaseen]

Nick Cox i have trade data over a period from 1995 to 2022...now i want to regress the tradd data over 3, 4 and 5 year intervals like i want to reg data over 3 year intervals like 1995, 1997, 2000 and so on, over 4 yr and 5 yr intervals in the same manner...how can i do this in stata...like what is the command to do this...thanks

[Nick Cox]

Sorry but I can't follow #8 well enough to give advice. For example, how is 1995, 1997, 2000 a 3 year interval? I would start a new thread with a clear data example.

[Dr. Iqra Yaseen]

Nick Cox by 3 year interval i mean if i start with 1995...i would skip 3 years and then again include 1999, then again skiping more 3 years i want to include 2003 in my regression equation...i want a simple and short command to do this in my reg command

[Nick Cox]

You've not started a new thread or given a data example and, sorry, I can't follow #10 any further than #8. If you want regression with rolling intervals, consider rolling (official), asreg (SSC) or rangestat (SSC). If you want regressions with disjoint intervals, consider statsby, by: or just using if conditions, as in

regress ... if inrange(year, 1995, 1997)

where the ... cover whatever else you want.

[Dr. Iqra Yaseen]

thanks Nick Cox