Prev: count for certain criteria.
Next: Out of memory Error for Unstructured correlation structure in GEE
From: saraliz80 on 14 Apr 2010 12:08
Hi,
I have an issue with one of my variables in my imputed dataset. It
seems it wasn't missing when I imputed it because I keep getting the
message: "Between-imputation variance is zero for variable" when I try
to combine analyses using proc mianalyze.

First of all, what does this error mean?

Second, I just need to report descriptives and run a regression. I get
an estimate of the mean using proc mianalyze anyway, but it does not
match the one I get when I just run proc means with the dataset. In
addition, I don't get a regression coefficient at all with proc
mianalyze. Should I just be using the original data to run this one
analysis instead of the entire one and trust those results?

Thanks-
Sara
From: Matt C on 15 Apr 2010 09:22
Hi Sara,
I can't say I know what the errors in your code mean but I have done a
little with MI and find it a more straight forward than MIANALYZE. I
have used MI followed by proc univariate for simple situations where
all I needed was the basic descriptives too. Try this on for size
maybe it will help.
Cheers,
Matt

*****************************************
OPTIONS NODATE;
TITLE1 'MCC: GRAVITY EXP. FOR MI';

DATA gravmiss;
INPUT R time height gravity;
DATALINES;
0.737 0.129 1.549 10.530
1.958 0.316 1.991 9.623
0.087 . 1.914 .
0.331 0.065 1.262 10.214
0.005 0.001 1.553 9.608
.. 0.081 1.609 .
0.284 0.051 1.671 9.391
0.805 0.151 1.538 9.205
1.719 0.291 1.916 9.132
.. 0.025 1.772 .
1.301 0.234 1.599 9.689
1.634 0.286 1.572 10.358
1.175 0.205 1.624 10.121
1.107 . 1.943 .
0.091 0.016 1.813 9.361
.. . 1.603 .
1.558 0.261 1.850 9.666
0.477 0.078 1.981 9.403
1.026 . 1.833 .
1.705 0.379 1.010 9.992
1.469 0.245 1.872 9.623
;

PROC PRINT DATA = gravmiss;
RUN;

PROC CONTENTS DATA = gravmiss;
RUN;

PROC MI DATA=gravmiss OUT=outmi NIMPUTE=1 SEED=37851;
TITLE2 '1st Proc MI';
VAR gravity R time height;
RUN;

PROC PRINT DATA=outmi;
TITLE2 'First-MI';
RUN;

PROC UNIVARIATE DATA=outmi;
VAR R time height gravity;
OUTPUT MEAN=gravity; * VAR=gravity CV=gravity;
RUN;

PROC PRINT DATA=outreg;
VAR R time height;
TITLE2 'Parameter Estimates from Imputed Data Sets';
RUN;
QUIT;
 | 
Pages: 1
Prev: count for certain criteria.
Next: Out of memory Error for Unstructured correlation structure in GEE