topic Re: PROC MIXED - Random statement uncertainty in Statistical Procedures

PROC MIXED - Random statement uncertainty

mabeg510 — Sun, 03 Jun 2018 02:18:04 GMT

I am analyzing a data set.

I have 3 x 4 x 4 design.

Three treatments with four doses for each treatment and four sample time points for each dose. The samples are individual samples at different times (not repeated samples).

I am using Proc Mixed as under

proc mixed;
class group dose time;
model e=dose|group|time/ outp=new;
random

data residy; set new;

proc plot;
plot resid*pred;
proc univariate plot normal;
var resid;

run;

I am not sure what should be my "random" statement.

I will appreciate any help. Thanks

Re: PROC MIXED - Random statement uncertainty

sld — Sun, 03 Jun 2018 03:40:28 GMT

If the observations are independent--if there is no clustering, or blocking, or repeated measures on the same subject--and if the homogeneity of variances assumption is met, then there is only one source of variance in your design (which will be estimated as residual variance) and you do not need a RANDOM statement. The MIXED procedure works just fine without a RANDOM statement.

Re: PROC MIXED - Random statement uncertainty

mabeg510 — Tue, 12 Jun 2018 19:52:18 GMT

Many thanks. It was of great help.