BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
njh1109
Calcite | Level 5

I am new to SAS and trying to run a PROC MIXED model. With one dataset, the model runs and gives me full output. For another dataset, the model seems to run, but the output only includes the following: 

- Model Information

- Class Level Information

- Dimensions

- Number of Observations

 

The rest of the output (i.e., "Iteration History," "Estimated G Matrix," "Estimated G Correlation Matrix," "Covariance Parameter Estimates," etc., etc.) do not show up.

 

HOWEVER, I am not getting any error or warning messages. 

 

Any thoughts on what the problem might be? 

 

Thanks so much!

 

1 ACCEPTED SOLUTION

Accepted Solutions
SteveDenham
Jade | Level 19

It's probaly the case that the spatial power covariance structure won't converge with so many different timepoints spread across the various subjects.  This is probably a very difficult problem, and I don't know if there is a good solution.  You might consider what was done in the documentation for PROC GLIMMIX Example 45.6 Radial Smoothing of Repeated Measures Data, and fit a low rank smoother (a structure not available in PROC MIXED).  What is going to make this difficult is that not all subjects are measured at all timepoints, so convergence may be a problem.  

 

Steve Denham

View solution in original post

8 REPLIES 8
ballardw
Super User

Since the data was different what differences did you make to the proc call. Output does change depending on options used. Without seeing the code and descriptions of the data used we would be guessing. Such as did your second model have any numeric variables?

njh1109
Calcite | Level 5

Thanks for your reply! Here is the PROC call. I am attaching a file with the Import Data output for both data sets--the one that is NOT working first, and then the one that IS working.

 

PROC MIXED covtest DATA=IMPORT METHOD= reml cl;
CLASS id;
MODEL actdis=timemin happyint lvl2 happyint*lvl2 happybtw happybtw*lvl2 /solution cl;
RANDOM intercept happyint/SUBJECT=id TYPE=un g gcorr cl;
REPEATED /SUBJECT=id TYPE=sp(pow)(timemin);

 

Thanks so much for your help.

ballardw
Super User

@njh1109 wrote:

Thanks for your reply! Here is the PROC call. I am attaching a file with the Import Data output for both data sets--the one that is NOT working first, and then the one that IS working.

 

PROC MIXED covtest DATA=IMPORT METHOD= reml cl;
CLASS id;
MODEL actdis=timemin happyint lvl2 happyint*lvl2 happybtw happybtw*lvl2 /solution cl;
RANDOM intercept happyint/SUBJECT=id TYPE=un g gcorr cl;
REPEATED /SUBJECT=id TYPE=sp(pow)(timemin);

 

Thanks so much for your help.


BOTH sets of code. Since the set that is working does not have any of the variables that start with HAPPY, Lvl2, Timemin or Actdis then the code above did not "work" for the working data set.

njh1109
Calcite | Level 5

Of course! Sorry about that. I'm attaching everything here.

 

Here is the code that IS working:

 

PROC MIXED covtest DATA=IMPORT1 METHOD= reml cl;

CLASS id;

MODEL intimacy=time7c confcw relqual confcw*relqual confcb confcb*relqual /solution cl;

RANDOM intercept confcw/SUBJECT=id TYPE=un g gcorr cl;

REPEATED /SUBJECT=id TYPE=sp(pow)(time);

 

and again, the code that it NOT working.

 

PROC MIXED covtest DATA=IMPORT METHOD= reml cl;
CLASS id;
MODEL actdis=timemin happyint lvl2 happyint*lvl2 happybtw happybtw*lvl2 /solution cl;
RANDOM intercept happyint/SUBJECT=id TYPE=un g gcorr cl;
REPEATED /SUBJECT=id TYPE=sp(pow)(timemin);

 

Thank you SO much for your time.

SteveDenham
Jade | Level 19

What is the repeated factor in each case?  My surprise is that one of these actually ran and gave meaningful results.

 

Steve Denham

njh1109
Calcite | Level 5

Both data sets are from intensive longitudinal studies in which individual participants were asked the same survey assessment questions repeatedly over a period of time. The repeated factor for both data sets represents time, and the first assessment/time point for each participant was set to 0. For the one whose code IS working (I'll call this Data Set 1), the time unit was days, and the range was 0 (i.e. the first day of the study/first assessment/time point) to 27. Some participants had missing days/time points.

 

For the data set whose code is NOT working (I'll call this one Data Set 2), the repeated factor was also time, but the units were in minutes. They are also a lot more spread out and variable than in Data Set 1. For example, whereas Data Set 1 participants all had 27 days of the study (even if they missed a day or two here or there in between), the participants for Data Set 2 were prompted randomly throughout the course of the day (over a 6 day period) to fill out the survey assessment. Although all participants were prompted an equal number of times per day for assessments, the particular analysis that I am currently trying to do is using only a subset of time points from the study (e.g., only time points for which the participant reported a certain type of stressful event). So, for example, the second time point for one participant might have been 30 (i.e., 30 minutes after the first time point) whereas for another participant it might be 240 (i.e., 4 hours after the first time point) or even 1440 (i.e., a full day after the first time point).

 

Dataset 1 & its code came from a book on working with intensive longitudinal data, so if one was going to work, I'm not surprised that it was that one!

 

Thanks so much for your time and thoughts,

Nastassia 

 

 

SteveDenham
Jade | Level 19

It's probaly the case that the spatial power covariance structure won't converge with so many different timepoints spread across the various subjects.  This is probably a very difficult problem, and I don't know if there is a good solution.  You might consider what was done in the documentation for PROC GLIMMIX Example 45.6 Radial Smoothing of Repeated Measures Data, and fit a low rank smoother (a structure not available in PROC MIXED).  What is going to make this difficult is that not all subjects are measured at all timepoints, so convergence may be a problem.  

 

Steve Denham

njh1109
Calcite | Level 5

Thank you so much for your feedback (and for bearing with me as I got you all of the information that you needed). I really appreciate it.

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
  • 8 replies
  • 3381 views
  • 0 likes
  • 3 in conversation