RE: st: egen with user-defined function

[Jan Barendregt <j.barendregt@sph.uq.edu.au>]

Hi Joe,

Thanks for your reply.
1) You're right, these are ICD10 codes. They are coded in a single string (see also my answer to Nick's reply, but, unfortunately, the way they are coded differes between years. And I also have a year with ICD-9 codes. So I'm trying to write something that with minor tweaks can handle all these variations. And yes, I'm at this point only interested whether a sepsis code is mentioned, not how many or which ones.
2) Using separate variables seems a pain, because there can be up to 20 different codes in the string.

Thanks,

Jan

------------------------------------------
Jan J Barendregt, MA, PhD
Assoc Prof of Epidemiological Modelling
School of Population Health, University of Queensland
Email: j.barendregt@sph.uq.edu.au
Skype: janbarendregt
Phone: +61 7 3102 3093
Visit www.epigear.com: home of Ersatz, MetaXL, DisMod II, Prevent, and more! 

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Joe Canner
> Sent: Tuesday, 20 August 2013 11:32 AM
> To: statalist@hsphsun2.harvard.edu
> Subject: RE: st: egen with user-defined function
> 
> A couple of minor comments:
> 
> 1. Based on the context and the fact that you named your 
> indicator variable "bool", I suspect you are only interested 
> in knowing whether codes in the range A410-A419 appear at 
> all, not how many there are.  In that case, Nick's code can 
> be simplified using the -strpos()- function to look for the 
> first occurrence of "A41x".  If this (as I suspect) is a 
> well-behaved list of ICD-10 codes, you could even just look 
> for "A41" and probably be OK.
> 
> 2. While Nick's code is, as usual, quite clever and quite 
> correct, I am wondering why you don't parse the 4-digit 
> (ICD-10?) codes into separate variables.  This, I believe, is 
> probably how they were meant to be processed and so doing 
> reduces the need for clever coding tricks.
> 
> Just my $0.02.
> 
> Joe Canner
> ________________________________________
> From: owner-statalist@hsphsun2.harvard.edu 
> [owner-statalist@hsphsun2.harvard.edu] on behalf of Nick Cox 
> [njcoxstata@gmail.com]
> Sent: Monday, August 19, 2013 8:40 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: Re: st: egen with user-defined function
> 
> I don't see quite what -egen- is objecting to here but you have more
> fundamental problems.
> 
> Wiring in particular names to a program is not illegal but it usually
> implies that you are pitching your problem at the wrong level.
> 
> (*) I guess that you have some variable -RECORD_AXIS_DATA- and you are
> looking for occurrences of the strings "A410" ... "A419".
> 
> Your code would at most look in the first observation as you are using
> the -if- command
> instead of an -if- qualifier.
> 
> It would be probably be best if you described the concrete problem, as
> my strong guess is that no original program is needed here.
> 
> But I'll guess that (*) is the nub of the problem.
> 
> clonevar copy = RECORD_AXIS_DATA
> 
> qui forval i = 0/9 {
>       replace copy = subinstr(copy, "A41`i'", "", .)
> }
> 
> gen count = (length(RECORD_AXIS_DATA) - length(copy)) / 4
> 
> So in a copy of the variable we replace each occurrence of "A410"
> "A411" ... "A419" with an empty string. Each such deletion reduces the
> number of characters by 4.
> 
> This was written up in 
> http://www.stata-journal.com/article.html?article=dm0056
> 
> 
> Nick
> njcoxstata@gmail.com
> 
> 
> On 20 August 2013 01:03, Jan Barendregt 
> <j.barendregt@sph.uq.edu.au> wrote:
> 
> > I'm trying to let egen to use a user-defined function. Here 
> is what I'm doing:
> >
> > do "C:\Ddrive\MattCooper\Data\test.do"
> >
> > . capture program drop _grecordparse
> >
> > . program define _grecordparse
> >   1. local bool=0
> >   2. local li=1
> >   3. local i=0
> >   4.   while `i' < RECORD_AXIS_COUNT {
> >   5.   if (substr(RECORD_AXIS_DATA,`li',4)>="A410") & 
> (substr(RECORD_AXIS_DATA,`li',4)<="A419") local bool=
> >> `bool'+1
> >   6.   local li=`li'+4
> >   7.   local i=`i'+1
> >   8.   }
> >   9.   gen `typlist' `varlist' = `bool'
> >  10. end
> >
> > .
> > end of do-file
> >
> > . egen sepsis=recordparse
> > unknown egen function recordparse()
> > r(133);
> >
> >
> > So I can't get egen to recognise my recordparse function. Any ideas?
> >
> > I'm using Stata 11.2 on Windows
> >
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
> 
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/