Proc FMM is not as flexible as some of the other procedures with respect to the CLASS statement and explicitly defining the reference level.  Instead, you can adjust the order of CLASS variable levels with the ORDER= option in the PROC FMM statement.   ... View more

This is to be expected.  Because the imputed values are a result of random draws from their respective distributions, changing the ordering of the variables will change where in the stream each distribution is being drawn from.  What should not change however is the posterior distribution from which those random numbers are drawn.  Think of it as similar to changing the seed.  You will get the same posterior distribution, but the draws from those distributions will differ. ... View more

I assume that the ERROR message you are receiving is the following: ERROR: The CLASS variables cannot be used as covariates in an FCS discriminant method with the default CLASSEFFECT=EXCLUDE option. The solution is to use the CLASSEFFECTS=INCLUDE option on the FCS DISCRIM statement. https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/statug/statug_mi_syntax05.htm#statug.mi.fcsdiscrimopt ... View more

I think most statisticians would question the usefulness of using a variable as a predictor that has 94% of missing values. That being said, if the goal is to use model selection, then multiple imputation isn’t really an option. Because it creates multiple imputed data sets it is probable that you will get different models selected for some of the data sets. This means you will not be able to combine the results for a single set of parameters. If you want to impute categorical variables, then you can use the DISCRIM or LOGISTIC methods on either the FCS or MONOTONE statements in Proc MI. This section of the documentation might be helpful: https://documentation.sas.com/doc/en/statug/15.2/statug_mi_details05.htm ... View more

This approach would also work for the relative risk with all of the CL= options other than EXACT. ... View more

Unfortunately, PROC FREQ doesn’t provide RISKDIFF analysis for two-way tables when there is a zero-frequency column(both treatment arms have no positive responses). In the meantime the work-around to assign very small weights to the zero-frequency cells in your two-way table is the only option. When you use this work-around, the following confidence limit types are available for the risk difference: Agresti-Caffo, Hauck-Anderson, Miettinen-Nurminen (score), and Newcombe. Of course the traditional/Wald confidence limits are also available, but equal (0,0) in the case of a zero-frequency column where the estimate of the risk difference is 0.    ... View more

You could try the suggestions in the usage note below. 24475 - Handling the ERROR: "An imputed variable value is not in the specified range after 100 tries" (sas.com) ... View more

There is a known issue in Proc MI that will be fixed in an upcoming release that requires you also include the MODELOBS= variable in the MIN/MAX= list.  It appends it to the beginning of the variable list on the VAR statement.  Try this code instead: proc mi data=widegrp1_2 seed=93538001 nimpute=50 out=newgrp1_2 noprint minimum= . . 0 0 maximum=. . 14 14 minmaxiter=600 ; class TRT01PN CSIRSTR ; fcs nbiter=50 reg(BASE = CSIRSTR); fcs nbiter=50 reg(W2 = CSIRSTR BASE); /*[MESURE]='DD'*/ MNAR model(W2/ modelobs=(TRT01PN = '3')); var CSIRSTR BASE W2; run;   ... View more

I suggest that you combine the LsMeans and standard errors from MIXED in MIANALYZE.  You could then save the p-values from the combined estimates to a SAS data set and feed them into MULTTEST.  You would have to use a different adjustment method than Tukey like Bonferroni since MULTTEST will not compute Tukey p-values from a data set of raw p-values. ... View more

The standard errors that I was referring to were associated with the point estimates themselves, not the combined estimates.  In other words, do you have a formula that reflects the standard errors that you will eventually average in the first half of Rubin's methodology? ... View more

In order to combine any statistics from multiply imputed data you need the point estimate and a standard error.  Do you know how the standard errors for standardized mean differences and synthetic Ns would be calculated? ... View more

PROC SURVEYREG uses elementwise regression to compute the regression coefficient estimators by generalized least squares estimation. The procedure assumes that the regression coefficients are the same across strata and primary sampling units (PSUs). To estimate the variance-covariance matrix for the regression coefficients, PROC SURVEYREG uses either the Taylor series (linearization) method or replication (like BRR) methods to estimate sampling errors of estimators, based on complex sample designs. ... View more

Proc MI does not have a specific option for dealing with count (poisson) data.  To circumvent this you could use the REGPMM method on a MONOTONE statement.  This will lead to only integer values being selected as imputed values. As far as the value of K= that you should select, there is not really an ideal value, but instead there is a tradeoff of sorts.  A smaller K= value tends to increase the correlation among the multiple imputations for the missing observation and results in a higher variability of point estimators in repeated sampling. On the other hand, a larger K= value tends to lessen the effect from the imputation model and results in biased estimators.  There is a good discussion in  Schenker and Taylor (1996 ) on  p. 430. ... View more