If the data are non-normal then what distribution would you use to describe them?  In this case it is likely that you would need to take a modeling approach if you are going to use SAS.  If the distribution comes from an exponential family then you could use a MACRO that is linked below that uses replicate weights in GENMOD. https://support.sas.com/rnd/app/stat/examples/SurveyPoisson/surveypoisson.htm   I don't see why you couldn't use different DIST= LINK= options for other distributions and link functions (you would need to modify the  MACRO code MODEL statement, but the idea should still hold.   ... View more

“Change point analysis” is a little bit of a vague term, but the following example from Proc MCMC might be a good place to start.  If this is what you ultimately want to do, then you would need to program it rather than use a point and click approach.   https://go.documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_mcmc_examples18.htm&locale=en ... View more

“Change point analysis” is a little bit of a vague term, but the following example from Proc MCMC might be a good place to start.  If this is what you ultimately want to do, then you would need to program it rather than use a point and click approach. https://go.documentation.sas.com/?docsetId=statug&docsetVersion=15.1&docsetTarget=statug_mcmc_examples18.htm&locale=en ... View more

I am not sure I understand what you mean by your comment about flipping the variables.  In general, your outcome variable should be the column, that is, the last variable on the TABLES statement, with the level of interest in the first column.  Your row variable would be ordered such that the row1 level will be compared against the row2 level in the odds ratio.   You should be able to control the ordering of the rows and columns by using the ORDER= option.  The easiest thing to do is to set up formats that properly reflect the ordering and use the ORDER=FORMATTED option. ... View more

I think this is the difference between the tests themselves.  The parameter estimates/LsMeans are tested using a Wald test while the Type3 test uses a Score test.  If you add the WALD option to the MODEL statement, then it will give you Wald statistics for Type 3 contrasts as well. ... View more

Because you have used the PROBMODEL statement, you are saying that the mixing probabilities depend upon the estimated parameters from those effects ( i.e. age, policy, gender, diag).  Since you only have 2 components, and the sum of the mixing probabilites is 1,  it is only necessary to know the estimates for the first component. Those estimates do not, per se, have anything to do with the lognormal component estimates.  They are simply estimates that enable you to calculate the mixing probabilties between the two components.    Can you be more specific about the predictions being odd? ... View more

I would say that first you probably should be using more than 5 imputed data sets, especially if your fraction of missing information is high (see the table of Relative Efficiency here.     This may make the results closer to what you are expecting, but quite frankly, your expectations are not correct.  If the combined estimates factored in only the within imputation variances, then it would be correct.  But it also includes the between imputation variance, which naturally will raise the variance of the estimates.  This is the price you pay for the uncertainty associated with imputation.  Therefore, you should expect that the number of significant estimates will be less (and often significantly less so) than any of the individual imputations. ... View more

MI is meant to impute based on a multivariate distribution and thus needs more than 1 variable. ... View more