turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- Hierarchical growth curve model using proc glimmix

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

07-18-2017 03:21 PM

Hi, I am trying to perform hierarchical age-period-cohort (HAPC) growth curve analysis using a longitudinal study. Data have been collected from more than 15000 individuals (aged 0-100 years at baseline) nine times at 2 years interval. Information on presence of asthma (binary) and age were collected at each data collection time point. Data set is in long format and sorted by ID. My outcome variable is asthma prevalence (binary). Age has been centered around the grand mean. I am also using quadratic age using the centered age variable. A cohort is defined by the birth year of an individual. sfwgt is the standardized survey weights for each individual. I want to assess if there is any intercohort differences in asthma prevalence and any intercohort variation in the age effects on asthma prevalence.

I am trying to follow the following HAPC growth curve model but using proc glimmix for my binary outcome:

proc mixed data=ACL_Depression covtest noclprint; class ID; model CES-D = Cohort Age_c Cohort*Age_c Age_c2 Cohort*Age_c2 Died Nonresponse /solution; random intercept Age_c Age_c2 / subject = ID type = un; weight sampleweight; run;

I started with a null model and gradually included variables in the fixed and random parts of the model. The model worked till I included quadratic age in the random part of the model using the following codes:

**proc** **glimmix** data=asthma maxopt=**25000** noclprint;

class id;

model asthma (descending)= cohort age_C cohort*age_C age_C2 cohort*age_C2 /solution dist=binary link=logit ddfm=bw;

random intercept age_C /subject=id;

covtest GLM/WALD;

NLOPTIONS TECHNIQUE=NRRIDG;

weight sfwgt;

**run**;

When I include quadratic age (age_C2) in the random part, the model did not converge. I tried using the following codes and get the error: The SAS System stopped processing this step because of insufficient memory:

**proc** **glimmix** data=asthma maxopt=**25000** noclprint;

class id;

model asthma (descending)= cohort age_C cohort*age_C age_C2 cohort*age_C2 /solution CL dist=binary;

random intercept age_C /subject=id;

covtest GLM/WALD;

NLOPTIONS TECHNIQUE=NRRIDG;

weight sfwgt;

**run**;

I also tried using TECHNIQUE=NEWRAP but get the same results. Since my data come from a complex longitudinal survey, should I use method=quadrature with proc glimmix statement?