topic proc logistic: seeking time-saving advice in Statistical Procedures

proc logistic: seeking time-saving advice

christine1 — Fri, 13 Oct 2017 15:23:34 GMT

hello.

i am running Proc Logistic on a dataset with 8 million observations. i am stratifying by calendar year (13 years. 2001-2013). the exposure and outcome are binary variables. i am adjusting by a propensity score (continuous number between 0 and 1. 9 decimal places)

proc logistic data=dsn;

strata calendarYear/ no summary;

model outcome (event = '1') = exposure ps / risklimits parmlabel;

run;

it is taking hours to run. any suggestions on what i can change to decrease processing time?

much gratitude.

christine

Re: proc logistic: seeking time-saving advice

Reeza — Fri, 13 Oct 2017 15:30:55 GMT

These would help for most SAS procs, but I'm not sure if they'll work for PROC LOGISTIC.

1. Do you have an index on your data set? If not consider adding it by calendarYear

2. You're stratifying your analysis. Is it a true 'stratification' or do you need separate analysis for each year? If the latter, then use BY and/or split your analysis into multiple steps.

3. Drop any/all other variables in the DATA statement since they're not used.

Re: proc logistic: seeking time-saving advice

Ksharp — Sat, 14 Oct 2017 12:45:02 GMT

Not sure if it could work.

proc logistic data=dsn;
strata calendarYear/ no summary;
model outcome (event = '1') = exposure ps / fast risklimits parmlabel;
run;