keep the row of information which is closest (in days) to a selected date variable

[Rose Matthews]

I would like to keep one row of information for each ID which has the

preop_date or recpreop_date CLOSEST to the surgery date

Please note I've posted this as a continuation of another thread, but I thought as this is another question sepeate to the original post I thought I would post separately

//Select which is the correct pre_opdate (from the received vs preopdate) by finding the one closest to the operationdate //Generates the days from surgerydate  

gen preop_correct1 = abs(operationdate-preop_date)  
gen preop_correct2 = abs(operationdate-recpreop_date)

 //I then don;t know how to proceed from here

Dataset:

* Example generated by -dataex-. For more info, type help dataex
clear
input float id str2 id2 float(pain scar surgdate oks gender preop_date recpreop_date postop_date recpostop_date operationdate) byte(_merge duplicatescores)
12 "1A" 1 1 22555 40 1 22533 22289 22770 22771 22555 3 0
13 "2B" 1 9 11962  1 2 11933 11933 11963 11782 11962 3 0
13 "2B" 1 9 11962 48 2 11932 11951 11963 11782 11962 3 1
13 "2B" 1 9 11962 48 2 11932     .     .     .     . . 1
end
format %td surgdate
format %td preop_date
format %td recpreop_date
format %td postop_date
format %td recpostop_date
format %td operationdate
label values _merge _merge
label def _merge 3 "Matched (3)", modify

[Girish Venkataraman]

There is a rowmin egen function you can use like this below. But you will have to have more assertion statements to make sure there are no missing values and differences are actually positives etc..Perhaps other expert Statalisters will nest this whole thing below is a single conditional statement.

egen wanted = rowmin(preop_correct1 preop_correct2)
bys id (wanted): keep if _n == [1] 
drop wanted preop_c*

[Rose Matthews]

Nice and simple ! Great one, was playing around with reshaping !