Adding Observations with a Fixed Time Value

[Nicole Moiseyev]

This is a sort of revision of my previous post:

https://www.statalist.org/forums/for...-time-variable

I figured this title was more descriptive of my issue and more open to different ways to solve it. To sum up that post, I need to add an observation with a precise Time value (which is in seconds for my dataset) for Time = 60, 120, 180, 240, and 300. I want the other variables in the new observations I create to match the last observation closest to the new time one. For example, I may have an observation at t=57.8 seconds and my next one is t=67.2 seconds. I want to add an observation for t=60 seconds and have all other variables in that observation match that of the t=57.8 observation.

Thank you so much in advance!

Nicole

[Clyde Schechter]

So something like this:

//  CREATE A DATA SET WITH JUST DESIRED VALUES OF
//  TIME, AND NOTHING ELSE
use original_data, clear
keep section
duplicates drop
expand 5
by section, sort: gen time = 60*_n
gen byte added_obs = 1

//  NOW BRING IN THE ORIGINAL DATA
append using original_data
replace added_obs = 0 if missing(added_obs)

//  SORT OBSERVATIONS ON TIME AND COPY
//  CLOSEST OBSERVATION BEFORE
ds time added_obs, not
foreach v of varlist `r(varlist)' {
    by section time, sort: replace `v' = `v'[_n-1] if added_obs
}

Not tested. Beware of typos.

Note: In providing this code I am not endorsing the wisdom of this approach. My instinct is that interpolation or some other approach might be more sensible than carrying forward, but that is a scientific issue that you and your collaborators would understand best.