Single-arm repeated measurement for continuous outcome / nonsymmetric data

Cruise — Fri, 15 Mar 2019 19:29:06 GMT

Hi SAS experts,

I'm trying to find out whether hemoglobin of the study subjects increased or decreased over time adjusting to agecat stratifying by race. So that I'll be able to tell the direction of trend by race controlling for the effect of agecat. There is no control group. Each subject has repeated measurements with no equal intervals. In other words, measurements taken whever they showed up at clinics. Number of repeated measurements also vary across subjects. For example, pat_id has 2 measurements while pat_id may have 6 measurements so forth so on.

My attempt for modeling this scenario using proc glimmix is shown below.

My questions are;
1. Does model in SAS code make sense given my research question?
2. Do you have any idea why final output doesn't provide P-value?

Thanks a lot in advance for taking your precious time for help.

The similar problem was solved in the forum below. However,my context is different. My endpoint is continuous not binary.

https://communities.sas.com/t5/Statistical-Procedures/Analyzing-a-single-arm-study-with-repeated-measures/m-p/174029#M8999

DATA HAVE;
INPUT PAT_ID	TIME	AGEcat	RACE	HB1;
CARDS;
1	1	5	2	13.5
1	2	4	2	12.8
2	1	4	2	11.5
2	2	5	2	11.9
3	1	2	1	11.9
3	2	3	1	12.8
4	1	5	1	13.5
4	2	3	1	11.8
4	3	4	1	11.1
5	1	5	1	11.1
5	2	5	1	12.5
6	1	5	1	11.3
6	2	4	1	12
7	1	4	1	12.5
7	2	5	1	12.5
8	1	4	1	11.5
8	2	5	1	12.7
9	1	5	1	12.8
9	2	4	1	13.3
10	1	5	1	11.8
10	2	3	1	11.6
11	1	5	1	14.2
11	2	4	1	14.2
12	1	5	1	12.1
12	2	4	1	12.4
13	1	4	1	11.8
13	2	3	1	10.7
14	1	4	1	11.6
14	2	3	1	11.8
15	1	4	1	12.8
15	2	3	1	12.5
16	1	4	1	13.7
16	2	4	1	13.4
17	1	5	1	13.3
17	2	4	1	10.8
18	1	4	1	13.1
18	2	5	1	11.7
19	1	3	1	12.8
19	2	4	1	12.3
20	1	3	1	12.5
20	2	4	1	13.2
21	1	4	1	13.3
21	2	5	1	12.9
21	3	3	1	12.1
22	1	3	1	12.9
22	2	5	1	11.2
22	3	2	1	11.9
23	1	3	1	12.8
23	2	4	1	11.8
24	1	3	1	12.8
24	2	5	1	12.2
25	1	3	1	11.2
25	2	5	1	11.4
26	1	4	1	11.6
26	2	4	1	11.4
27	1	5	1	11.8
27	2	4	1	12.2
28	1	4	1	11.1
28	2	5	1	11.6
28	3	3	1	12.1
29	1	3	1	12
29	2	5	1	11.5
30	1	4	1	12.9
30	2	5	1	13.5
31	1	4	9	10.3
31	2	4	9	12.7
32	1	4	3	10.8
32	2	5	3	10.8
33	1	5	1	12.3
33	2	4	1	12
34	1	5	1	12.3
34	2	5	1	12.2
35	1	4	1	12.4
35	2	5	1	12.4
36	1	4	1	13.2
36	2	5	1	10.1
37	1	4	1	12.4
37	2	3	1	14.4
38	1	4	1	10.5
38	2	3	1	12.4
39	1	5	1	13.8
39	2	5	1	12
40	1	5	1	12.9
40	2	3	1	13.1
41	1	4	1	12.7
41	2	4	1	12.3
42	1	2	3	12.2
42	2	4	3	11.9
42	3	4	3	11.1
43	1	2	1	12.8
43	2	3	1	13.3
44	1	5	1	11.6
44	2	4	1	11.8
45	1	5	3	14.2
45	2	4	3	12.6
46	1	4	1	11.6
46	2	5	1	13.6
47	1	5	1	12.7
47	2	3	1	11.1
;
PROC PRINT; RUN; 

proc glimmix data=HAVE;
   class TIME PAT_ID RACE;
   model HB1 = TIME RACE TIME*RACE/solution;
   random intercept / subject=PAT_ID;
   random intercept / subject=PAT_ID(TIME) type=UN;
   LSMEANS TIME RACE TIME*RACE/DIFF; 
run;

Re: Single-arm repeated measurement for continuous outcome / nonsymmetric data

Norman21 — Sat, 23 Mar 2019 20:48:49 GMT

Your program generates a warning message; possible solutions are discussed from the bottom of page 13 of the following document:

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

Re: Single-arm repeated measurement for continuous outcome / nonsymmetric data

Cruise — Wed, 27 Mar 2019 19:39:50 GMT

@Norman21

Thanks Norman. What also helped me was the SUGI paper in the link below.

https://support.sas.com/resources/papers/proceedings/proceedings/sugi27/p253-27.pdf

topic Re: Single-arm repeated measurement for continuous outcome / nonsymmetric data in Statistical Procedures

Single-arm repeated measurement for continuous outcome / nonsymmetric data

Re: Single-arm repeated measurement for continuous outcome / nonsymmetric data

Re: Single-arm repeated measurement for continuous outcome / nonsymmetric data