Blups equivalent to Empirical Bayes mean estimation for the random effects for the stata command mixed ??

[Jian Zhang]

Dear Statalisters,

I was trying to predict the random effects after fitting a mixed-effects linear regression model using command mixed. The manual says that using command predict after command mixed will give the best linear unbiased predictions (BLUPs) of the random effects. On the other hand, the manual for command meglm (a command for estimating a mixed-effects generalized linear model) says that command predict after meglm gives the empirical bayes mean estimation and the empirical bayes mode estimation of random effects when remeans and remodes are specified in the option, respectively. I was just wondering whether the best linear unbiased predictions (BLUPs) of the random effects in command mixed are empirical bayes mean estimation? The manual for mixed postestimation does not say the BLUPs is the empirical bayes mean estimation.

In addition, my understanding is that since the command mixed assumes a normal distribution of error terms, the empirical bayes mean and mode should be same as a result of the normality of the distribution. Is this correct?

Thanks!

Jian

[Jian Zhang]

anyone knows the answer or know what paper should i read about this?

[Carlo Lazzaro]

Jian
may want to take a look at Rabe-Hesketh S, Skrondal A. Multilevel and Longitudinal Modeling Using Stata. 3rd edition. Volume I, page: 109-115 (especially). College Station, TX: Stata Press, 2012. (please, see for further information: http://www.stata.com/bookstore/multi...ata/index.html ).

[Jian Zhang]

Thanks, Carlo. It turns out that BLUPs is Empirical Bayes Mean estimation using predict after mixed. See the thread: http://www.stata.com/statalist/archi.../msg00868.html
In the book you provide, it also says so, see page 82. Thanks.

Jian

[Carlo Lazzaro]

Jian:
thanks for closing this thread.
Happy with reading that both literature and Statalist helped you out.