st: xtlogit

[David Airey <david.airey@vanderbilt.edu>]

Previously, I posted:

> I have N white bags and N black bags filled with 1000 marbles each. 
> Marbles comes in 10 colors. There is bag to bag variation in the 
> proportion of each color marble in a bag, and an unknown white and 
> black bag treatment effect on the distribution of proportions of 
> marbles. I only get to sample X marbles per bag.
>
> We have a biological problem that is analogous to this. Scientists 
> have analyzed these data by ignoring bag variation and just focusing 
> on numbers of marbles sampled from white versus black (equivalent to 
> pouring all black bags into one black bag, and all white bags into one 
> white bag).
>
> I think this is not a good idea, but what is an appropriate Stata 
> estimator that accounts for the bag to bag random effect, the white 
> and black fixed effect, and models the count/categorical/proportion of 
> marbles as dependent variable?
and got no response. I got some response off list. I just want to say I 
think xtlogit or xtgee now seems appropriate. If I simplify the 
situation to the case where there are just black and white marbles 
mixed in each of the black and white bags, we have a binary dependent 
variable (marbles) sampled from clusters (bags).

Does anybody have a favorite reference to share where the scientist who 
makes the mistake of ignoring the cluster level and the consequences 
for inference on the treatment (bag color) effect? I'm assuming the 
consequences are loss of control over alpha, but is the direction 
predictable?

-Dave

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/