[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: vce(cluster ID) or collapse?

From   Andrea Bennett <[email protected]>
To   [email protected]
Subject   st: vce(cluster ID) or collapse?
Date   Fri, 8 Aug 2008 21:31:01 +0200

Dear Statalisters,

I have a data set on the individual level containing knowledge answers related to multiple topics, e.g.

ID1 age education answer (on topic1) interest
ID1 age education answer (on topic2) interest
ID2 age education answer (on topic3) interest
ID2 age education answer (on topic4) interest
ID3 age education answer (on topic1) interest
ID3 age education answer (on topic2) interest

In the first scenario, the answers (dependent variable) are explained by multiple other independent variables, e.g. the probability to correctly answer knowledge related questions for multiple topics is explained (while it is not the case that every individual has to answer questions related to the very same topics; I include topic specific dummies here).
However, when I switch to another scenario where I use a dependent variable which does not vary on the individual level (e.g. "general interest in politics" in the above example), would I then have to - collapse- the data set -by(ID)- or would it still be o.k. to use the - vce(cluster ID)- option in the regression? I think I should collapse the data set here. However, there might be situations where I would like to explain -interest- by data related to the different topics, which would then be completely impossible.

I might quite generally be a little confused applying the -vce- option. E.g. when I merge regional data into the above data set (which is on the individual level), would I then principally have to cluster on these regional levels (e.g. states), too? AFAIK, I could use - gllamm- to cluster on multiple levels when using logti/probit which is the case here.

Any clarification on this issue would greatly help my understanding!

Kind regards,


P.S. if there is a possibility to collapse all variables except the one specified, that would make applying the -collapse- command a lot easier to handle since I got about 20 something independent variables. I understand I cannot use -collapse _all, by(ID)-.

* For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index