Regular Learner
Posts: 1

# Multiple Imputation: NIMPUTE = and which PROC to use in Step 2

I'm at an intermediate SAS level, but I'm trying to use PROC MI and PROC MIANALYZE for only the first time.  I believe I'm using SAS 9.2.  I've searched websites and texts for answers to my questions but can't find anything clear and definitive.

My data has 23 variables with missing value rates from 0% up to 25%.  The initial step, as follows,works in reducing the arbitrary MV pattern to a monotone missing pattern:

/* Use MCMC to reduce the data set with an arbitrary MV pattern to */
/* a monotone missing pattern, from 185 to 10 missing patterns */

PROC MI DATA = Anxiety.Transformed NIMPUTE = 25 SEED = 19530125 OUT = MonotonePattern
MIN = 1,0,-3.083,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
MAX = 99999.9999,16,+2.821,4.3,2.5,6,3.5,36,20,3.9,4,3,3,3,4,3,4,36,15,256,4.5,4,6
ROUND = .0001,1,.001,.001,.001,.001,.001,1,1,.001,.001,.001,.001,.001,.001,.001,.001,1,1,1,.001,.001,.001;
MCMC IMPUTE = MONOTONE;
VAR AWTCW01L ABECS08 AINHD08 sqrtASPHS01 sqrtASFHS5 sqrtAFNHS01 sqrtASFHS7 MatDep ABECS07 sqrtASFHS6
sqrtAC1CS01 sqrtBHTCbS1b
;
RUN;

All variables are continuous.  I know that after the above I should run PROC MI, some sort of analysis, and PROC MIANALYZE.  However my overall analysis plan is to impute all MVs, perform factor analysis, combine variables into factors, and only then run regressions on the factors.

I don't believe I can use regression analysis in Step 2, between PROC MI and PROC MIANALYZE, because I have six outcomes, but also at this stage I am simply trying to replace missing values.

Question: What sort of analysis should I be running in Step 2?

Question: What should I specify as the value for NIMPUTE = at any step?