model building question

From: alison@c255.ucsf.EDU (ABoeckmann) Subject: model building question Date: 19 Aug 1997 18:42:44 -0400 Attached is a question that was sent today by Paul Laub. Full text of the question is attached below. My comments are marked with >. QUESTION 1: Is my interpretation correct? >Yes. QUESTION 2: Is there still some way of getting individual ETAs at this early stage of model development? >Yes, use Exponential moel for eta, as you suggested. >This is suggested in Guide VII, Conditional Estimation Methods, >Chapter III, p. 8. QUESTION 3: How do I explain the equality of ETAs returned from the exponential and proportional representation? >This is not so surprising. Guide VII, above, tells you to expect just >what you have seen. Please look through Guide VII - it is only 12 >pages long and contains much valuable information about posthoc and >other conditional methods. Finally I would like to gain so physical intuition about the meaning of individual ETAs and elements of the OMEGA matrix. I can readily do this for the additive and proportional representations but have no idea how to do this for the exponential representation. QUESTION 4: Can anyone help me here? >With first order estimation method, there is no difference between >proportional and exponental models for eta. The interpretation of >OMEGA is the same. In Guide IV, Chapt. 3, it is stated: > >These estimates are empirical Bayesian esti- mates, conditional not >only on the data, but, importantly, also on values for the >population parameters. If the first-order estimation method is >used, they may be obtained after the population parameter >estimates have themselves been obtained. > >I don't know what "physical intuition" you are looking for. Many >physiological characteristics (e.g., body weight) have skewed rather >than symmetric distributions in the population, so it is not such >an odd way of modelling them. >One further note: > >Guide V was for the most part written prior to the developement >of the conditional estimation methods, and before people realized >how useful posthoc etas etc. can be early in model development. >The short course content has changed since it was written. >That is why the additive/proportional models for eta appear in >Guide V rather than the (under some circumstances) equivalent >exponential model. Sorry if it you found it misleading in that >respect. Alison Boeckmann