re: st: understanding the -asis()- option in -stsplit-

["Jesper Lindhardsen" <JESLIN01@geh.regionh.dk>]

Hi there,

Even though it seems that I am alone in my troubles with understanding the -asis()- option in -stsplit-, I thought others may have the same difficulties in the future. In the essence, I confused spells and individuals. Here is an extract of my Statatech correspondence:


--→
I understand your concern. Actually, the  command -stsplit- is not intended to split on a variable that varies within individual. The option -asis()- works like more -force- options, just bypassing the checkings that the variable should be constant within individual. Notice that splitting is always performed within individual, not within spell. That is, -stsplit- will go back to the spell where the time to split is, within individual. I realize that our documentation could be more clear on this aspect, and I'll make a note about this.

The solution we provided (me: use -stset- on spells, split and then stset on individuals again) would work for your particular case, and it could be generalized to many cases by using -stset- with  copies of your "underscore" variables. Let me know if you want more examples on that.

<-----


HTH (someone),

Jesper



Jesper Lindhardsen
MD, PhD candidate
Department of Cardiovascular Research
Copenhagen University Hospital, Gentofte
Denmark





--------------------------
Original post:

Hi All,

This may be simple, but I am having a hard time with a futher split (-stsplit-) of multiepisode survival data (Stata 11.2/MP on windows server).
Basically I want to split some of the observations within an individual, if exposed, into several new observations. However, this (I figure) causes the offset (-after()- option) to vary within individuals, which forces the use of the -asis()- option-. But I cannot figure out the proper use, and the documentation on this option is scarse.

In the below very simplified example, I want to split the episodes with changevar==1 to be further divided in two, however, this does not work (only one of two episode is splitted).
I’ve tried several other stsplit commands, but no succes. 



clear all

input time	id	changevar	event

30	1	.	0

40	1	1	0

50	1	.	0

60	1	1	0
			
end

stset time, id(id) f(event)

stsplit hej if changevar==1,at(2) after(time=asis(_t0))

sort time


Hope you can help, 

Jesper


Jesper Lindhardsen
MD, PhD candidate
Department of Cardiovascular Research
Copenhagen University Hospital, Gentofte Denmark


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/