Local macro in "forvalues" function

[Guest]

My first post here, but have used this forum extensively in the past to trouble shoot. However, can't find a solution to my stata problem, if there is one.

I have a series of variables from DX1, DX2, DX3 ... DX25.

Variables DX1 - DX25 have a diagnostic code, which is a string value.

I'm trying to recreate a new variable that searches DX1-25, and pulls the diagnostic codes I am interested in.

Write now, I have this code:

gen AKI=0
forvalues j=1/25{
replace AKI=1 if inlist(DX`j', "5845", "5846", "5847", "5848", "5849")
}

But the problem is, for the inlist function, I can't include more than 5-10 string variables in the "inlist" function. For some new variables I am trying to generate, it will be based on 100s of diagnostic codes. So I thought if I use a macro with my string variables predefined, then I could avoid the 5-10 limit.

I've tried creating the following macro and incorporating it into my loop:

local aki "5845" "5846" "5847" "5848" "5849"

gen AKI=0

forvalues j=1/25{
replace AKI=1 if inlist(DX`j', "`x'")
}

And it doesn't work. The result is "0 real changes made"

Any help you might have would be tremendously appreciated. Thank you.

[Susan Wilson]

The below code worked fine. Perhaps recheck you initial try for typos or little errors.

clear
set obs 50
gen counter = _n
gen DX1=""
replace DX1 = "5845" if counter==45
gen DX2=""
replace DX2 = "5848" if counter ==3
gen DX3 = ""
replace DX3 ="5849" if counter == 27


gen AKI=0
forvalues j=1/3{
replace AKI=1 if inlist(DX`j', "5845", "5846", "5847", "5848", "5849")
}

[Susan Wilson]

Sorry! I just reread why you're interested in using a local variable. I'll give it a another look.

[Guest]

Thank you for your response. They are coded as string variables.

My question was how to create a macro with my observations of interest (i.e. diagnostic codes) for variables DX1 - DX25, and to use that macro in my "inlist" function, so I can avoid the 5-10 string variable limit.

I had thought of doing:


local x "5845" "5846" "5847" "5848" "5849"

gen AKI=0

forvalues j=1/25{
replace AKI=1 if inlist(DX`j', "`x'")
}


But the above produces "0 observations".

Thank you.

[Susan Wilson]

There are two problems: 1, using a local variable won't work with 'inlist', because it requires commas. 2, the local variable needs to be inside the loop.

What about something like this, which just uses a nested loop. Although it might be a bit slow in large numbers.

clear
set obs 50
gen counter = _n
gen DX1=""
replace DX1 = "5845" if counter==45
gen DX2=""
replace DX2 = "5848" if counter ==3
gen DX3 = ""
replace DX3 ="5849" if counter == 27

gen AKI=0
forvalues j=1/3{
forvalues k = 5845/5849 {
replace AKI=1 if DX`j'=="`k'"
}
}

[Guest]

Thank you so much for your response.

But what if k is for a string variable? As in, 5845 - 5849 are stored as string variables. I was interested in knowing how to apply the above with a string value.

[William Lisowski]

From post #1

So I thought if I use a macro with my string variables predefined, then I could avoid the 5-10 limit

That is the source of your misunderstanding. Stata will first expand all the macros in a command with their values, and then count the number of arguments after all macros have been replaced by their expansions. So using a local macro does not avoid the problem.