[R-sig-ME] LME4 to MCMCglmm

[tw at netstorm.be]

I figured the random effect should look like this:

~us(1+var1+var2):study

but now I am struggling to define a proper prior. Again, any help would be
welcome




2012/5/4 <tw at netstorm.be>

    Hi all,

    I am trying to run an lme4 model (logistic regression with mixed effects)
    in MCMCglmm but am unsure how to implement it properly.

    Currently, my lme4 model formula looks as follows: "outcome ~ (1 + var1 +
    var2 | study) + var1 + var2"

    In English, this means that I am fitting a random effects model, where
the
    intercept, var1 and var2 are jointly distributed according to study.

    My question is now how I would translate this formula to the fixed and
    random terms in MCMCglmm.

    For the fixed part, I figured that I should make a variable
    nooutcome=abs(1-outcome) because it can then be modeled with a
    multinomial2 family as there is no binomial(logit) option available.
    Then, the fixed part would look as follows:

    cbind(outcome,nooutcome)~1+var1+var2

    However, I am unsure how to specify the random effects over the
intercept,
    var1 and var2 jointly. So far, I was able to generate the following:

    random=~us(var1):study+us(var2):study+us(1):study

    which I think corresponds to  "outcome ~ (1  | study) + (0+var1  | study)
    + (0+var2  | study) +  var1 + var2" instead of "outcome ~ (1 + var1 +
var2
    | study) + var1 + var2"

    I would appreciate any help.

    Thomas

    _______________________________________________
    R-sig-mixed-models at r-project.org mailing list
    https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models