BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
sanketshah07
Obsidian | Level 7

Hello,

 

I am trying to measure the difference in change of eGFR (postindex_egfr) between treated (cohort_num=1) and untreated (cohort_num=0) groups. I am using a repeated measures mixed model to study this association and eGFR values have been divided by quarter (over 5 years). A lot of quarters will have missing eGFR values.

 

There are no duplicate values by quarter for any given patient. The code is written below and I'm getting an infinite likelihood error. Advice would be appreciated.

 

proc mixed data=hcvrenal.analysis_set_3 method=reml covtest;

 

class patid cohort_num (ref='0') gender bus_num dm(ref='0') htn(ref='0') ccf(ref='0') mi(ref='0') pvd(ref='0') cva(ref='0') can(ref='0') abuse(ref='0')
hiv hbv ftg(ref='0') ckd_stage(ref='1') quarter (ref='1');


model postindex_egfr = cohort_num ageatindex gender bus_num
dm htn ccf mi pvd cva can abuse hiv hbv ftg
fib4 ckd_stage
QUARTER indexdate  /s ddfm=satterth COVB ;

 

WHERE QUARTER IN (1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20);

 

RANDOM patid/cl;

 

repeated quarter/type=ar(1) subject=patid*cohort_num ;
lsmeans cohort_num/cl;

 

estimate 'Change from baseline to quarter 4' quarter
-1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0/e cl;

 

estimate 'Change from baseline to final eGFR' quarter
-1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1/e cl;
run;

1 ACCEPTED SOLUTION

Accepted Solutions
sld
Rhodochrosite | Level 12 sld
Rhodochrosite | Level 12

There could be some helpful tips here

http://support.sas.com/resources/papers/proceedings12/332-2012.pdf

and

http://support.sas.com/resources/papers/proceedings15/SAS1919-2015.pdf

 

I have found that rescaling the response can be miraculous (sometimes anyway).

You may have collinearity problems with the predictor variables (both continuous and categorical).

 

You could start with a basic model that includes only a minimal number of predictors (i.e., cohort_num and quarter). If that converges, then you have something to work with. If it doesn't then that may point you towards a potential problem with patid.

View solution in original post

4 REPLIES 4
Shmuel
Garnet | Level 18

I'm not familiar with this procedure but you missed posting the ERROR you got.

Please post, too, the log and any results you got, marking the errors.

sanketshah07
Obsidian | Level 7

Here is the log:

 

WARNING: Class levels for patid are not printed because of excessive size.

 

NOTE: 20364 observations are not included because of missing values.

 

WARNING: ODS graphics with more than 5000 points have been suppressed. Use the PLOTS(MAXPOINTS= )
option in the PROC MIXED statement to change or override the cutoff.

 

WARNING: Stopped because of infinite likelihood.

 

 

Shmuel
Garnet | Level 18

I have googled for:  sas proc mixed infinite likelihood

 

it seems to me that next link may help:

https://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_mixed_sect0...

 

Next lines were copied from that link page:

An infinite likelihood during the iteration process means that the Newton-Raphson algorithm has stepped into a region where either the  or  matrix is nonpositive definite. This is usually no cause for concern as long as iterations continue. If PROC MIXED stops because of an infinite likelihood, recheck your model to make sure that no observations from the same subject are producing identical rows in  or  and that you have enough data to estimate the particular covariance structure you have selected. Any time that the final estimated likelihood is infinite, subsequent results should be interpreted with caution.

 

 

You may understand it better than me.

sld
Rhodochrosite | Level 12 sld
Rhodochrosite | Level 12

There could be some helpful tips here

http://support.sas.com/resources/papers/proceedings12/332-2012.pdf

and

http://support.sas.com/resources/papers/proceedings15/SAS1919-2015.pdf

 

I have found that rescaling the response can be miraculous (sometimes anyway).

You may have collinearity problems with the predictor variables (both continuous and categorical).

 

You could start with a basic model that includes only a minimal number of predictors (i.e., cohort_num and quarter). If that converges, then you have something to work with. If it doesn't then that may point you towards a potential problem with patid.

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 4 replies
  • 12654 views
  • 0 likes
  • 3 in conversation