I am using proc glimmix in SAS to fit a multilevel model for a multinomial outcome with unordered response categories. Proc glimmix requires that you specify the group= option in the random statement to obtain random effects for each outcome group. However, I suspect this prevents that correlations between random effects of different outcome groups are computed, even if the type=un option is specified.
Is there any way within the glimmix procedure to circumvent this issue? Thanks in advance!
I don't think there is any way to get Proc Glimmix to do this directly. I don't know if SAS constrains the parameters to be independent between groups, so a rudimentary way around the problem might be something like saving the random effects to a dataset and the use Proc Corr to get the correlation/covariance of the random effects between groups.
That might be completely wrong, but that is just my intuition!
I do not know about SAS but the resources on multilevel unordered multinomial at
http://www.bristol.ac.uk/cmm/software/mlwin/mlwin-resources.html#discrete
might help
Thank you professor Jones. I already read those sources, they are exellent. Actually, I was one of your students in the multilevel course for the QASS program in Brussels several years ago. I am specifically looking into proc glimmix right now, because I want to extend a SAS macro I wrote previously.
Great
I am told that SAS are very helpful in responding to technical queries
Good luck!
Kelvyn
Randomness seems so contrary to correlations. If a correlation can suggest randomness of outcome from groups, the correlation may as well imply randomness of outcomes from groups. A multivectorial correlation between randomized outcomes and randomized groups is itself random. Hence a correlation of randomness implies randomness. A good mathematician can prove the statement. I reserve my intuition that randomness and correlation are the antonyms in Oxford Statistics Dictionary.
Kumar
It is perfectly possible to have correlated random effects. The term Random Variable simply means 'allowed to vary' - that is variables can take on different values (according to a distribution) and does not necessarily mean 'independent'.
Laure,
I looked into the following sources to find meaning of random (After the web-links, follows the statistical meaning).
- http://www.oxforddictionaries.com/definition/english/random: Governed by or involving equal chances for each item.
- http://www.thefreedictionary.com/random: Of or relating to a type of circumstance or event that is described by a probability distribution, Of or relating to an event in which all outcomes are equally likely, as in the testing of a blood sample for the presence of a substance.
Once a sample of blood has been tested, the blood sample can hardly take on different values according to a probability distribution.The probability distribution is obtained after blood samples are tested. I have yet to see an inverse scenario. Perhaps the possibility of effect of randomness can bear a correlation with random outcomes. A graph of random data poses a challenge for several statisticians to fit any meaningful curve of correlation.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics