Covariance structure for a repeated measure data

Emma_at_SAS — Fri, 18 Dec 2020 19:16:14 GMT

I have repeated measures data with 3 measurements for which the correlation between repeated measures is shown below. I use PROC MIXED and have /TYPE= in the REPEATED statement.

I tried different covariance structures but only the model with UN(1) converges. I was expecting that AR(1) would converge because I thought that would be a better fit based on the PROC CORR results. I appreciate it if I have your insight on this.

(I tried ARH(1), AR(1), UN, VC, CS, UNR, TOEP, CSH, TOEPH)

proc corr data=wide;
var M1 M2 M3;
run;

Pearson Correlation Coefficients Prob > \|r\| under H0: Rho=0 Number of Observations
	M1	M2	M3
M1	1.00000 42167	0.85104 <.0001 41466	0.78900 <.0001 40845
M2	0.85104 <.0001 41466	1.00000 41923	0.81367 <.0001 41028
M3	0.78900 <.0001 40845	0.81367 <.0001 41028	1.00000 41446

Thank you!

Re: Covariance structure for a repeated measure data

SteveDenham — Mon, 21 Dec 2020 14:37:11 GMT

What sort of error messages, warnings or notes are in the log? What about messages in the output listing? Some of those might help shed some light on what is happening. I find it surprising that UN(1) converges and VC doesn't, so let's see what is happening there first.

SteveDenham

topic Re: Covariance structure for a repeated measure data in Statistical Procedures

Covariance structure for a repeated measure data

Re: Covariance structure for a repeated measure data