topic Re: proc glimmix help in Statistical Procedures

proc glimmix help

nbryan5_ — Tue, 17 Jun 2014 19:15:02 GMT

I am attempting the procedure in sas with the following code, but SAS keeps hanging up. Naturally I assume I have done something wrong with the statement:

data counts;

infile "/folders/myfolders/counts2.txt" dlm='09'x missover lrecl=10000 firstobs=2;

input flight $ cells;

proc print;

run;

proc sort data=counts out=counts2;

by flight;

run;

proc glimmix data=counts2;

class flight;

model cells=flight link=log dist=negbinomial;

run;

What am I missing?

Re: proc glimmix help

Reeza — Tue, 17 Jun 2014 19:19:39 GMT

For starters where is it "hanging up"?

Run each piece individually and check the log in between to identify the issue. And how big is your dataset (counts2) with how many distinct flights?

Re: proc glimmix help

nbryan5_ — Tue, 17 Jun 2014 19:23:14 GMT

I am running SAS studio and when I click the running man I getting a running...... message with a spinning wheel but no conclusion, and no error log.

There are about 3000 lines in the data set from 10 different flights.

Re: proc glimmix help

Reeza — Tue, 17 Jun 2014 19:28:43 GMT

Separate it out. Run the import first and see if that works. If that means copying the code over reach time then do so. I haven't played around with SAS Studio so don't know if there's a submit selected or submit all button yet.

For example, does this alone work?

data counts;

infile "/folders/myfolders/counts2.txt" dlm='09'x missover lrecl=10000 firstobs=2;

input flight $ cells;

run;

Re: proc glimmix help

nbryan5_ — Tue, 17 Jun 2014 19:48:12 GMT

gave sas studio a reboot and I now have output.

I can even add the lsmeans and get an answer. Is there a way to incorporate a Tukey style comparison in the glimmix?

Re: proc glimmix help

SteveDenham — Wed, 18 Jun 2014 15:16:20 GMT

There is a way, but I would suggest using the Edwards/Berry method (simulate= option). Tukey's method depends pretty strongly on a gaussian distribution for accurate adjustments. (See Westfall et al., Multiple Comparisons and Multiple Tests Using SAS, 2nd ed.). The code would look like:

lsmeans flight/diff adjust=simulate(seed=1) adjdfe=row;

Steve Denham