"using data not sorted" when using stptime

[Abdul-Kareem Abdul-Rahman]

Dear Statalisters,

I am trying to generate Standardised mortality ratios (SMRs) by alcohol-use cohort groups.

I am using Stata/SE14.1.

my dataset looks like this.

* Example generated by -dataex-. To install: ssc install dataex
clear
input long patid byte(gender cohort) float indexdate int exitdate byte(death _st _d) double(_origin _t) byte _t0
20767109 1 4 18079.32 20426 0 1 0      18079.3203125  6.425034737432921 0
27783206 1 4 17349.32 20446 0 1 0      17349.3203125  8.478479048023218 0
11762154 1 4  5296.32 12075 0 1 0   5296.31982421875 18.559522987025655 0
 2611322 1 3 -3104.68 20447 0 1 0 -3104.679931640625  64.48275088062815 0
55991525 1 4 18079.32 20439 0 1 0      18079.3203125  6.460627772149819 0
 9637473 1 2 13331.32 20439 0 1 0      13331.3203125 19.460299221060126 0
22185415 1 1 17714.32 18197 0 1 0      17714.3203125 1.3215411441791698 0
 3041518 1 3 15888.32 19015 0 1 0      15888.3203125   8.56061682044683 0
 7268556 1 4  1643.32 19785 0 1 0 1643.3199462890625 49.670572921122925 0
12960427 1 3 13696.32 18415 0 1 0      13696.3203125 12.919394610393166 0
end
format %td indexdate
format %td exitdate
label values gender genderlab
label def genderlab 1 "male", modify
label values cohort cohortlab
label def cohortlab 1 "lifetime-abstainer", modify
label def cohortlab 2 "low-risk", modify
label def cohortlab 3 "hazardous", modify
label def cohortlab 4 "AUD", modify
label values death yesnolab
label def yesnolab 0 "no", modify

I am trying to run what I hope could be a way to generate SMRs by cohort group using the command

stptime, by(cohort) per(1000) smr(age popdist) using(analysis/eurostandpop)

the eurostandpop is the european standard population 2013 (http://ec.europa.eu/eurostat/documen...d-5e8fa09b3f8f)

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte age int pop double popdist
15 5500  .06547619047619048
20 6000  .07142857142857143
25 6000  .07142857142857143
30 6500  .07738095238095238
35 7000  .08333333333333333
40 7000  .08333333333333333
45 7000  .08333333333333333
50 7000  .08333333333333333
55 6500  .07738095238095238
60 6000  .07142857142857143
65 5500  .06547619047619048
70 5000  .05952380952380952
75 4000 .047619047619047616
80 2500  .02976190476190476
85 1500 .017857142857142856
90  800 .009523809523809525
95  200 .002380952380952381
end

However, this command seems to return the error

using data not sorted
r(5);

Would appreciate any help with helping me obtain SMRs for the alcohol cohort groups with stptime by solving the above or any other suitable advice or commands.

Thank you in advance for your kind help.

Yours truly,

Kareem

[Clyde Schechter]

OK, you have stumbled over a quasi-bug in -stptime-. The syntax error arises when -stptime- attempts to use the -merge- command, due to conflicting expectations about versioning.

In modern versions of Stata, -merge- does not require the data sets involved to be pre-sorted: the sorting is done behind the scenes if needed. But -stptime- dates back to (at least) version 7, and begins by specifying version 7 syntax as governing within its scope. Back in version 7, data sets had to be separately sorted before you could -merge- them. When you use the version 7 syntax (that is, -merge keyvars using using_file- as opposed to modern -merge 1:1 keyvars usng using_file-), the old requirement for sorting is resurrected. It looks like the code for -stptime- hasn't been modified to account for newer usage.

As a workaround for you, before you run -stptime- sort the -using()- data file on your -by()- variable and I think it will run without problems.

StataCorp may want to update the code in -stptime- so that when run on a later version of Stata, the pre-sorting is no longer needed.