Variables could include standard demographics like age, gender, disease history, or medication status.  What strikes me as an interesting addition to these are genomic variables--presence or absence of any number of genes in the individual.  All of this could be fed into a clustering algorithm. Steve Denham ... View more

Other than getting out the X'X matrix, and then doing a whole boatload of programming, I don't see how to do it in SURVEYREG.  I wonder if the observation-wise deletion even makes sense in a survey setting, so that it is just very difficult to interpret.  What occurs when you look at the results from PROC REG, as a guide to identifying influential observations? Steve Denham ... View more

See.  I knew I would learn something new on this.  So, those look like reasonable random statements at this point. Everything looks in order.  Since everything is G-side, I would select METHOD=QUAD in the PROC statement, but that is a matter of preference, and like a lot of things in GLIMMIX, may not be available for a nominal multinomial. Do you plan on calculating specific odds ratios?  I only ask because the default would be calculated at the mean of landscape1 and the mean of landscape2, and it sounds to me like you may want to investigate the odds ratios at different values of these endpoints.  I think that is a subject for a different day and the ESTIMATE statement. Steve Denham. ... View more

Not sure if I have the design correct, but: The two variables (landscapevariable1 and 2) are continuous measures that you use to predict response.  That strikes me as very unusual for a multinomial--you have no classes here, so GLIMMIX will fit these as continuous covariates.  More information on these would be extremely useful. YEAR stands alone as a random effect.  I assume that the families and individuals observed in any one YEAR are different from those in all other YEARs.  If not, then you may have a repeated effect, and I don't want to go there just yet. INDIVIDUAL is always nested within FAMILY. RESPONSE is a dependent variable.  It should probably not be included in the CLASS statement. So, what about two random statements: proc glimmix data=data; class individual family year; model response = landscapevariable1 landscapevariable2 / dist=mult link=glogit; random intercept family/subject=individual; random year/subject=individual(family); Now, about the GROUP= option in the random statements.  This may be an opportunity for me to learn something.  I have never seen grouping by the dependent variable, so I excluded it in my code.  It looks like you want to accommodate heterogeneity of variance for the dependent outcomes. The more I consider this, the more likely I think it is that I have misinterpreted the design. Is it possible that the landscapevariables are the actual dependent variables? Waiting for more info. Steve Denham ... View more

Ksharp's mention of McNemar's test is right on.  I don't know if PROC FREQ gives an exact test for McNemar's test, but with small sample size, it is certainly worth investigating. Steve Denham ... View more

Ksharp is trying to get you on the right track here.  If this is time series data, getting a whole bunch of t values, say one for each time point, and then averaging them falls into the category of a very bad idea.  First, the values are not independent and have some sort of serial correlation. Second, how could you possibly calculate the correct degrees of freedom to come up with a p value or standardized effect size?  I think you need to look at the ETS suite, especially PROC PANEL, if this describes your dataset. Try this for a start--look at PROC MIXED for a repeated measures analysis, with country and time as CLASS variables.  Your model should include the interaction between these to get separate estimates by country and time.  I don't know what the repeated nature of the time variable is like, but an AR(1) structure probably fits if your data are equally spaced in time.  It may be heterogeneous by country, so that is something else you will have to investigate.  In any case, the Type III tests are "averaged t-tests" in this context.  Specific t values can be obtained with an LSMEANS statement. Good luck. Steve Denham ... View more

Here I sit as SGF2012, so I can't pull up the docs that I would if I were in the office.  What might work: get output datasets from SURVEYREG (not sure if there is an output statement, but as always ODS OUTPUT is your friend).  Then it is off to PROC UNIVARIATE for Q-Q plots, or any of the SAS/GRAPH procedures (SGPLOT comes to mind), and you can plot the values obtained. Hope this helps some. Steve Denham ... View more

I think Jeff has hit the nail on the head--not enough data to estimate the random variability.  Another possibility is that the maximum likelihood estimate is actually negative, so when the standard REML estimate is computed, it comes out as a zero.  Negative estimates of random effects are almost always due to pathologies arising from insufficient data. Steve Denham ... View more

I know GLIMMIX can do the joint modeling of variables--follow along the example in the documentation for the binary/poisson, and replace poisson throughout.  Probably have to fix up the dataset so that the variable 'dist' is on each record. I don't think you can do it in GENMOD, but GLIMMIX should give the same results, assuming that you have no random effects.  If you are doing repeated measures in GENMOD (GEE stuff), you can get approximately similar answers in GLIMMIX with an appropriate RANDOM statement with the residual option. Steve Denham ... View more

Hi Renee, I don't know if Doc did things the same way I did, but I have the same issues. OLD: Had a list of subscribed communities.  Log-in, go to them directly.  Unread posts would appear in bolded font, indicating that there were new threads or responses to threads that I could go to immediately. If everything in a community was read, or was not something I really needed, I would "Mark All as Read." NEW; Log-in. Go to Places tab, go to Following flagpoint, start down the line, guessing whether I have read all of the posts in those threads I recall, and not being able to distinguish all new content.  When I finish with one community, I return to Places, go to the Following flagpoint, Repeat.  I cannot mark anything as read, so on my next visit, I have to go through a ton of material that I have already looked at. Even 20 years ago in Usenet user-groups, there was the ability to mark a thread or a post as read.  For the life of me, I don't understand why Jive doesn't have this capability. Thanks, Steve Denham ... View more

I think the problem may be that plots are identically numbered across sites, so that there is still replication at the subject level of plot.  Have you tried subject=plot*site? You might also consider whether to specify site as a random effect, rather than as a fixed effect.  That all depends on your inference space assumptions about site, and you would need to consider the interactions with site as probable error terms for testing the treatment effects. A third consideration would be defining the subplots as 1 through 4, and then extracting tests about the subplot effect using LSMESTIMATE statements. Hope this triggers some ideas. Steve Denham ... View more