Hey everyone - there is a way to use PROC MIXED to fit multiple response variables. A good example can be found in the documentation for the TYPE= option for the REPEATED statement. Look at TYPE=UN@AR(1), where Height and Weight for children measured over time is given. The code suggested is:   proc mixed; class Var Year Child; model Y = Var Year Var*Year; repeated Var Year / type=un@ar(1) subject=Child; run; A similar situation is included in the PROC GLIMMIX documentation, but is much more difficult to implement. Start with the example: Joint Modeling of Binary and Count Data. Note that this approach requires different distribution types for each of the dependent variables, which is not required for the PROC MIXED method.   SteveDenham   ... View more

I am not sure this will work but you might try writing ESTIMATE statements that combine the solutions for the fixed and random effects. All I can suggest is to start simple and work it on up.   SteveDenham ... View more

This is just a gut feeling - the product of the coefficients and the weights makes sense to me, but I cannot regenerate your weights given only the sample size. Could you share how you get a weight vector of [1.9828 1.9574 1.6015 2.0167] from [4 4 3 4} for the linear contrast, and [1.9917 2.3803 1.7002 1.8362] for the quadratic?   SteveDenham   There should be an easier way than this.   SteveDenham ... View more

I think you are, but you should be aware that the variable "assists" probably isn't normally distributed, nor are the residuals of the model you are trying to fit. You will likely need a generalized linear model, and maybe the repeated nature can be modeled as a random repeated effect in PROC GLIMMIX.  Try this:   Proc glimmix data=dsn; class home_away player; model assists= home_away/ solution dist=poisson; random Intercept/subject=player; random home_away/residual subject=player; run; You possibly could use PROC GEE, which might give more interpretable results. I'll call on @StatDave who knows a lot more about generalized linear model analysis than I do. Here would be my attempt:   proc gee data = dsn; class home_away player; model assists= home_away / dist=poisson link=log ; repeated subject = player / type=exch covb corrw; run; SteveDenham       ... View more

This behavior may be due to misspecifying the "block" variable for Variants=1 and 8 (Block 3 appears twice, with no Block 4) and Variants=4 and 5 (Block 4 appears twice, with no Block 3). Another complicating factor is the imbalance in the data, where treatments 5 and 6 only were measured at fertilizer level = 0.  Combining these may make your mapping to the trtXX variable such that you have overlap where there are zero estimates or confounded estimates.   Have you tried fitting a model with only Variants 1 and 2? That may shed some light on what is going on.   SteveDenham ... View more

Some general thoughts regarding the modeling of binomial data, some of which might apply here.   First, using a generalized linear (LOGISTIC, GENMOD) or generalized linear mixed model (GLIMMIX) does not require either homogeneous variance or a normal distribution of residuals. Why not homogeneous variance? Because the variance can be directly expressed as a function of the mean, so if the group means differ, then group variances differ. No way around that on the original observed scale.  What about normality? Look at the model being fit. There is no additional variance term as in a linear model.  Now that does not mean that the data might not be overdispersed where additional variability above that due to the mean is present, but still that extra variability is not required to be gaussian (normal).  Thus the usual checks for assumptions in a linear model aren't quite as appropriate.   But that does not mean that examination of the residuals is a wasted effort. It can help you evaluate model appropriateness, or aid in checking the distributional assumption.  I hope this helps. If you are going to continue to fit models to binomial responses, get a good text (or on-line text) that covers the assumptions. Hosmer and Lemeshow's Applied Logistic Regression, McCullagh and Nelder's  Generalized Linear Models or Stroup's Generalized Linear Mixed Models are three really good sources, in my opinion.   SteveDenham would be good   ... View more

Consider changing the last RANDOM statement to:   random intercept/subject=study_id; Having study_id as both the random effect and the subject of the random effect most likely is the source of your error.   SteveDenham   ... View more

Thanks @FreelanceReinh !   SteveDenham ... View more

I would suggest the following code:   proc glimmix data=mydata plots=residualpanel method=laplace; /* you might want to explore method=quad or method=fastquad */ class group condition time; model Looking= group condition time group*condition group*condition*time /dist=beta link=logit ddfm=kr2 solution; random intercept / subject=ID type=un; random time / subject=ID(group*condition) type=un; run; So long as time is a class variable, a random intercept for only time isn't going to be meaningful, especially when there are only two levels. The GLM parameterization sets the last level to zero in the solution. I think the change in red in the last RANDOM statement will address the issue you are describing. If it doesn't, then you might consider dropping the first RANDOM statement completely and using your existing second RANDOM statement as the only one.   Other changes are to shift from the pseudo-likelihood default method to METHOD=laplace, and to use the Kenward-Roger2 method for ddfm, as it builds in a shrinkage based approach to the standard errors.   SteveDenham   ... View more

In addition to @Ksharp 's observation about model complexity, I would suggest that you look into using the NLOPTIONS statement in your GLIMMIX code. If the iteration history is relatively well behaved (i.e., it is close to monotonic throughout), simply increasing the number of iterations from the default of 20 using the MAXITER= option may result in convergence. I almost always set this at MAXITER=1000, especially when fitting binary data.   SteveDenham ... View more

I realize this is late, but this is one situation where some of the options in GLIMMIX come in handy (see the GLIMMIX documentation). You can test for homogeneity of variance using the COVTEST statement in conjunction with using the Group= option in the RANDOM statement. My advice would be to fit the heterogeneous model no matter the result of the test, AS LONG AS you have sufficient data to fit the  number of parameters needed for the covariance structures under consideration and the model converges without messages or errors.   SteveDenham ... View more

Would there ever be a reason to not include DIFF=method>? Absolutely yes. All of the methods are designed to control type I error (false positives). However, suppose that it is far more important to control type ii error (false negative rate) in whatever field you are working. Then the unadjusted values give a greater ability to declare a difference significant. As an example, suppose you are working on a safety study for a new drug. In that study, it would be much more important to detect a true difference to protect you from declaring the drug "safe" when in fact it causes some change that may indicate a potential safety concern.   SteveDenham ... View more

On your ESTIMATE statements, I assume you wish to find marginal values over Type, Treatment and the Type by Treatment interaction, as there are no terms included in the statement for those factors.   Have you considered using LSMESTIMATE statements, with the AT= option for various interesting values for the covariate?   SteveDenham ... View more

With regard to GLIMMIX, I was incomplete in my assertion about including a categorical response in the CLASS statement for multinomial distributions. That only seems to be the case for a generalized logit (LINK=GLOGIT), but not for the default link for DIST=MULT, where the default is a cumulative logit.   SteveDenham (Apologies to @StatDave ) ... View more