Normal or gamma random effects?

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Dear list,

For mortality data, children (level 1) within families (level 2), I would like to compare the fit of a logistic regr. model using normally distributed random effects with a model using a skewed random effect distribution, like a gamma distribution. What software could you recommend me to make such comparison? Thanks for any help,

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：effects Effect normal random gamma normal

Comment re SAS only:

A logit-normal assumption is easily addressed in GLIMMIX and, with a little more effort, NLMIXED.  Also MCMC.  At least as of a couple years ago or so, SAS didn't have an easy way of fitting a binomial-gamma mixture in a likelihood setting.  But see

• Nelson KP, SR Lipsitz, GM Fitzmaurice, J Ibrahim, M Parzen and R Strawderman. 2006. Use of the probability integral transformation to fit nonlinear mixed-effects models with nonnormal random effects. J Computational Graphical Statistics 15: 39-57.
• NLMIXED and MCMC code for fitting both models may in principal be obtained by eliminating zero-inflation terms from code at http://www.umesc.usgs.gov/staff/bios/bgray/code/zibsas.html
  

Thanks for pointing me to SAS and the paper. It's probably not as simple/common as I thought. Also there's the hglm package in R, which seems to enable gamma-distributed random effects, but I'm not sure what "h-likelihood" means and whether it could be used to compare models with different random effect distribution. So, may be I'll have to purchase a new license for SAS. Best regards,