Conformability error, r(503)

[Karishma DSouza]

Hi. I've been getting the conformability error r(503) on running the code below. Obviously my matrix specifications are off, but I can't seem to figure out where. Any help would be much appreciated. Thank you.

local time "r30 r90 r365"
local total "total"
local base_char "ib3.stratumn i.dschrgyr"  
local pati_char "ib1.racecat1n_mice c.age##c.age ib1.stkcat1n ib1.sscat2n_mice ib2.sexn ib0.hisstktian"
local hosp_char "ib2.hrucacat2n ib2.medschcatn"
local other_char "ib2.rucacat2n_mice ib0.dc_refneedn_mice ib0.dc_refhhn_mice ib0.pcpfln ib2.dualcatn ib0.medhisdpn ib0.medhishtn ib0.medhissmn ib0.anycvdn"


*******************************
* coefplots for total payments
*******************************

cap program drop multireg
program multireg
syntax varlist(numeric) [if]
local timestamp=subinstr("`c(current_time)'", ":", "_", .)
local indvar i.tx
*four specifications
local spec1 ib3.stratumn i.dschrgyr
local spec2 ib3.stratumn i.dschrgyr ib1.racecat1n_mice c.age##c.age ib1.stkcat1n ib1.sscat2n_mice ib2.sexn ib0.hisstktian

*placeholder for results
mat m = J(9,2,0)
loc b = 6   
foreach i in `*' {
loc n = 1
while `n' <= 2 {
display " "
display "This is a regression of `i' on `indvar' with `spec`n''."


**********************************************
**     30-day model -   TPM (gamma)    **
**********************************************
if "`i'" == "r30_total_w"   {
tpm `i' `indvar' `spec`n'', first(logit) second(glm, fam(gamma) link(log)) cluster(hospclus)
margins, at(tx=(0 1))
mat a = r(table)  
sca a1 = a[1,1]  
sca a2 = a[1,2]  
mat m[`b', 1] = a1, a2
loc b = `b' + 1
margins, dydx(tx) post
est store m`i'_`n'
outreg2 `*' using "./table/`i'_`timestamp'.doc", ctitle("`i'_`n'") alpha(.01, .05, .1)
}


******************************************************************
** 90-day model - TPM(inverse guassian) **
******************************************************************
else if "`i'" == "r90_total_w" {
tpm `i' `indvar' `spec`n'', first(logit) second(glm, fam(igaussian) link(log)) cluster(hospclus)
margins, at(tx=(0 1))
mat a = r(table)
sca a1 = a[1,1]
sca a2 = a[1,2]
mat m[`b', 1] = a1, a2
loc b = `b' + 1
margins, dydx(tx) post
est store m`i'_`n'
outreg2 `*' using "./table/`i'_`timestamp'.doc", ctitle("`i'_`n'") alpha(.01, .05, .1)
}


**********************************************
** 365-day model - GLM (gamma)    **
**********************************************
else if  "`i'" == "r365_total_w" {
glm `i' `indvar' `spec`n'', fam(gamma) link(log) cluster(hospclus)
margins, at(tx=(0 1))
mat a = r(table)
sca a1 = a[1,1]
sca a2 = a[1,2]
mat m[`b', 1] = a1, a2
loc b = `b' + 1
margins, dydx(tx) post
est store m`i'_`n'
outreg2 `*' using "./table/`i'_`timestamp'.doc", ctitle("`i'_`n'") alpha(.01, .05, .1)
loc n = `n' + 1
}
loc b = `b' + 1
}
putexcel set "$output/table/coefplot.xlsx", sheet(Total Payments) replace
*rownames
matrix rownames m = "0" "0" "30-day" "Base Covariates" "+Patient Covariates" "90-day" "Base Covariates" "+Patient Covariates" "365-day" "Base Covariates" "+Patient Covariates" 
*main table
putexcel A1 = mat(m), nformat(number) rownames
mat list m


**********************************************
** format table     **
**********************************************
*title
putexcel A1 = "Post-acute total payments for COMPASS patients versus control patients by timewindows"
*subtitles
putexcel B2 = "Adjusted mean spending", bold
putexcel A4 = "Spending", bold border(bottom)
putexcel B3 = "Control", bold
putexcel C3 = "Treatment", bold
putexcel B4 = "N=201", bold border(bottom)
putexcel C4 = "N=140", bold border(bottom)
putexcel D4 = "Differences in Spending", bold border(bottom)
putexcel (A1:I1), merge
putexcel (B2:C2), merge hcenter
putexcel (D4:I4), merge hcenter
end

multireg r30_total_w r90_total_w r365_total_w

*generate coefplots
coefplot ///
(mr30_total_w_?, label("30-day")) (mr90_total_w_?, label("90-day")) (mr365_total_w_?, label("365-day")), ///
keep(1.tx) aseq swapnames xline(0) ///
mlabel(cond(@pval<.001, "$"+string(@b, "%9.0f")+"***", cond(@pval<.01, "$"+string(@b, "%9.0f")+"**", cond(@pval<.05, "$"+string(@b, "%9.0f")+"*", cond(@pval<.1, "$"+string(@b, "%9.0f")+"+", "$"+string(@b, "%9.0f")))))) mlabposition(1) mlabsize(medium) ///
yscale(off) groups(mr30_total_w_? = "{bf:30-day}" mr90_total_w_? = "{bf:90-day}" mr365_total_w_? = "{bf:365-day}", nolab) ///
coeflabels(*1 = "base" *2 = "patient" *3 = "hospital" *4 = "other") legend(off) grid(none) ///
note("+ p < .1, * p < .05, ** p < .01, *** p < .001", size(vsmall))

graph export "./graph/totpmt_adjusted.png", replace

*put into excel (to combine with other statistics)
putexcel set "$output/table/coefplot.xlsx", sheet(Total Payments) modify
putexcel D5 = picture("$output/graph/totpmt_adjusted.png")


log close

[Leonardo Guizzetti]

This post contains a lot of code that is specific to your project, which we cannot work with in Stata. As asked, you are unlikely to get a helpful response. By now, you should be familiar with the FAQ and how to ask questions here.

Could you at least point which line is causing the conformability error? At a guess, you might be trying to access or assign to an index that does not exist in your -m- matrix.

[Karishma DSouza]

Leonardo Guizzetti You are right, that code is very specific to the project. If I could've broken it down and provided it here, I would've. Unfortunately, I am to adopt someone else's code (who has left) from a shared project to my part of the project, and a large part of the problem lies in there.

Since the loop is so long and has multiple nested loops within, it's hard to identify what exactly is driving the problem here. The code breaks down after the following lines, so your hunch on the -m- matrix seems right. Now to figure how to resolve it.

 margins, dydx(tx) post  
est store m`i'_`n'

[Leonardo Guizzetti]

Your example is still unworkable so my generic advice then is to use

set trace on
set tracedepth 1

To try to watch your code run as macro variables are substituted. You can turn trace off again when finished. Pay attention to the dimensions of m and how you index it.

[William Lisowski]

Here is my guess.

You loop over your three arguments, and for each pass through the loop you do

mat m[`b', 1] = a1, a2
loc b = `b' + 1

The local macro starts at 6, but at the end of the loop over your arguments, you have

loc b = `b' + 1
}

So the first time through the loop, b will be 6 for the matrix assignment. The second time through the loop, b will be 8. The third time through the loop, b will be 10. But your matrix x has only 9 rows.

If this is the case, you could have easily found this by including

display "b = `b'"

before each of the matrix m assignments.

If you don't have enough information to solve a programming problem, as yourself, "how can I get more information to help me find the problem?" Rather than "how can I deduce an answer with the information I have?"