Here is a good place to get started. 24447 - Examples of writing CONTRAST and ESTIMATE statements ... View more

A short summary is impossible. I highly recommend you read the book SAS for Mixed Models, 2nd edition (2006), by Littell et al. ... View more

You are getting t statistics with that option. That is, you are seeing the pairwise difference of estimated expected values divided by the estimated standard error of the difference. This has a Student t distribution under the null hypothesis of no difference. I think Paige's point is that if you pulled out just the observations for one set of factor levels (throwing out the rest), and did a traditional t test, you would not get the same t statistic (this is true). Nevertheless, you are looking at t statistics with the MIXED output. It is debatable which adjustment for multiplicity that one should use. There is now a lot of support to use the adjust=simulate option, because this is based on fewer assumptions. ... View more

Do not transform ... View more

Oops, I wasn't paying attention that you have 0s. This transformation is only defined for nonzero positive values. You could add a small constant to all the data, but this becomes ad hoc (but commonly done). You might want to resort to nonparametric methods. You could use either of the NO tests you mentions. ... View more

The reference is the SAS/STAT User's Guide. It was example 6 in an older version of SAS. This example has been replaced by a more complex example in SAS 9.3 and later: SAS/STAT(R) 9.3 User's Guide This is more complex than you need. ... View more

Log transformation would definitely not work (opposite effect from what you need). You could try the Box-Cox transformation. THis can be done with PROC TRANSREG (see example 6). If the data for the two groups are stacked, with g=0 for the first group and g=1 for the second, then you would use: model boxcox(y) = identity(g); ... View more

For 1, it is actually fairly common for the covariance to be negative. I get this result a lot for some data structures and models. You are not correct that "the covariance is always positive". If subject is a block, then a negative covariance means that two randomly chosen observations within the same block are negatively correlated (or that there is more variation within blocks than between blocks). And for repeated measures, it is definitely possible for observations within a subject to have a negative correlation. Can't answer your second question because the link is too general. Not clear what part you are asking about. But B is generic for main effect B (random effect) and AB is the interaction of A (presumed fixed) and B. I highly recommend that you get a copy of SAS for Mixed Models, second edition (2006; SAS, Inc.), by Littell et al. ... View more

I would need to see what they did, because the IG is undefined at 0. One could model the data as a mixture of two distributions, with a different distribution for zeros. But you can only do this in Genmod with discrete distributions. Perhaps PROC FMM. ... View more

Citing a particular chapter, section, or page number in the User's Guide isn't automatically helpful because there are different versions of SAS in use, and the chapter number and page number are different in the relevant User's Guide for each version. ... View more

You can't. The inverse gaussian distribution is only defined for nonzero positive values. Same for the gamma. ... View more

TRANSREG has a MONOTONE transformation, but I have never used it. Check out the User's Guide and also different sections of: http://www.soc.iastate.edu/Sapp/soc512Kuhfeld.pdf ... View more