Here is the log detail for  %put &=SYSVLONG;SYSVLONG=9.04.01M3P062415 ... View more

Hi Garnett, Many thanks. I have got a very clear explanation and solution. However, I came up with another problem while running my actual data. I have got 11 (Level 3: subject = region) and 622 (Level 2: subject = V11) subject effects. Is there any limit on the maximum number of subjects? How do the subject effects affect the convergence? Moreover, is there a limit on the number of fixed effects and random effects on the model building process? For instance, I have got 15 - 20 fixed effects and 7 - 10 random effects on a three-level mixed effect model that I will be identifying which variables have an effect on the outcome of interest. The computer is taking too long to finish, most of the cases, it does not finish and ends up with error messages. Regards Teketo ... View more

Hi Garnett,   Many thanks. I got it clear now. However, I have ambiguities about how to specify the random intercepts and slopes for three level hierarchical model in Proc MCMC. Lets assume that we have 4 individual level variables (V01, V02, V03, V04), 3 variables at cluster level (V05, V06, V07) and 2 region level variables (V08, V09). In PROC GLIMMIX model, I can build the model like this;   Proc glimmix data = care; Class cluster region; model y (event = last) = V01 V02 V03 V04 V05 V06 V07 V08 V09 / S; random int V01 V02 V03 V04                   V05 V06 V07 / Subject = region; *7 random slopes and a random intercept; random int V01 V02 V03 V04 / Subject = cluster; *4 random slopes and a random intercept; run;   Here is my attempt in PROC MCMC; Data xcare; V11 = compress (cluster) || "_" || compress(region); Run;   Proc mcmc data = xcare seed = 570 nmc = 100000 outpost = mcmc_care; Parms B0-B9; *Since I have nine fixed effects as it is built in proc glimmix; Parms S2g S2d; Prior B: ~ normal (0, var = 1e6); Prior S2g ~ igamma (shape = 0.01, sacle = 0.01); Prior S2d ~ igamma (shape = 0.01, sacle = 0.01); random gamma? ~ normal (0, var = S2g) subject = region; *How can I specify the 7 random slopes and random intercept? ; random delta? ~ normal (0, var = S2d) subject = V11; *How can I specify the 4 random slopes and random intercept? ; Mu = B0 + B1*V01 + B2*V02 + B3*V03 + B4*V04 + B5*V05 + B6*V06 + B7*V07 + B8*V08 + B9*V09; P = logistic (Mu + gamma? + delta?); Model y ~ binary (P); Run;     How can I specify the same model like proc glimmix with seven random slopes and a random intercept for the Subject = region and the four random slopes and a random intercept for the subject = V11?   Regards   ... View more

Hi Garnett,   Many thanks. Does subject = clusterXRegion in proc mcmc has the same meaning as subject = cluster (region) that I used in proc glimmix? Does it mean cluster is nested within region, which I am interested in?   Regards ... View more

Hello, Many thanks for the details, this will help me in managing the classification variables. Unfortunately, this doesn't answer my main concern, how to build the random effects. I checked through SAS documentation; however, I couldn't find any showing nested random intercept and random slope models like what I put in the glimmix procedure. Regards ... View more

Hello,   Many thanks.   I did include all the class variables in the model statement. The SUBJECTs in the RANDOM statement are discrete, i.e. REGION ranges from 1 to 11 and V001 from 1 to 622. SAS stop processing the procedure; I am getting the following error message:   {  proc glimmix data = cs.caesarean2 method = laplace; class region hregion; model M17_1 (event = last) = v012 v106 v717 v501 v701 v705 v150h v136 v190 v130 v743a v201 m14_1 m10_1 v212 FACTYPE distance csa1 csa2 csr1 csr2 gr1 gr2 hfma v025 region HREGION/s cl dist = binary link = logit; random intercept v012 v106 v717 v501 v701 v705 v150h v136 v190 v130 v743a v201 m14_1 m10_1 v212 FACTYPE distance csa1 csa2 csr1 csr2 gr1 gr2 hfma v025 /subject = region cl s type = vc; random intercept v012 v106 v717 v501 v701 v705 v150h v136 v190 v130 v743a v201 m14_1 m10_1 v212 /subject = v001 (region) cl s type = vc; covtest / wald; run;   NOTE: The GLIMMIX procedure is modeling the probability that M17_1='1'. ERROR: The SAS System stopped processing this step because of insufficient memory. NOTE: PROCEDURE GLIMMIX used (Total process time):       real time           7.40 seconds       cpu time            6.61 seconds }   Kind regards Teketo ... View more

Dear Reeza,   Thank you indeed. The join can’t be made at the cluster level. We only have region as a common variable. Both datasets (the fictitious and the main datasets) have multiple clusters and health facilities per region. Every region has both clusters and health facilities; none of the regions missed either of the two.   Here are the datasets and the codes used.   *To calculate the distance; Data clusters; Input Cluster Lat1 Lon1 region; datalines; 1 14.0 38.7 1 2 11.6 41.5 2 3 12.7 37.5 3 4 7.5 39.1 4 5 12.7 37.6 1 6 10.4 40.1 2 7 12.1 40.0 2 8 12.2 37.0 3 9 9.9 39.9 3 10 9.1 40.7 4 ;   *Health facility data; Data health_facility; Input ID Lat2 Lon2 region HFMA FPRS; Datalines; 1 13.6 39.1 1 1 0 2 13.3 38.9 1 2 1 3 10.7 40.1 2 1 1 4 10.3 40.2 2 1 1 5 11.4 40.7 2 1 1 6 10.5 39.5 3 2 0 7 11.6 39.6 3 1 1 8 11.3 37.9 3 2 1 9 9.1 35.6 4 2 0 10 9.5 34.8 4 1 1 11 14.3 37.8 1 2 0 12 13.5 39.4 1 1 1 13 10.4 40.2 2 2 0 14 9.2 40.3 2 1 1 15 9.9 39.8 3 2 1 16 11.0 37.1 3 1 1 17 10.6 39.4 3 2 0 18 8.5 34.5 4 2 1 19 7.2 38.2 4 1 1 20 8.8 35.8 4 2 1 ; run;   *Proc sql: Using cross join to merge to the two datasets; proc sql; create table dist_table as select h.id, h.lat2 as lat_facility, h.lon2 as long_facility,      h.region,      c.cluster, c.lat1 as lat_cluster, c.lon1 as long_cluster,      geodist(h.lat2, h.lon2, c.lat1, c.lon1, 'k' ) as dist_km from clusters as c cross join health_facility as h where c.region=h.region order by h.id, c.cluster, dist_km; quit;     proc sort data = dist_table; by id dist_km; run;   *BY group and FIRST; Data merged; Set dist_table; By id dist_km; If first.id; Run;       Kind regards Teketo   ... View more

Dear Reeza,   Thank you indeed. The dataset has health facilities in each region; however, all the clusters are not included in the final dataset. Below in the table, I put the number of health facilities and clusters along with the respective regions given in the fictitious dataset.   Table showing the number of clusters and health facilities in each region Region Number of Health facilities Number of Clusters 1 4 2 2 5 3 3 6 3 4 5 2 Total 20 10   I used the health facility id and distance to sort the dataset and in the BY group and FIRST data step too. However, it didn’t include all the clusters; for instance, clusters 3, 5, 7 and 10 were not included. I am not getting the desired final dataset.   Could there be any other possible option?   proc sort data = dist_table; by id dist_km; run;   *BY group and FIRST; Data merged; Set dist_table; By id dist_km; If first.id; Run;   Kind regards Teketo ... View more