panel data - generate variable that shows variation

[Katharina Meyer]

Hey everyone,

for my bachelor thesis i am doing a logit regression. i have a panel dataset. the survey was conducted 5 times.
as i tried to do a fixed effects regression i realized that there isn't enough variation in y as almost all observations had to be dropped because of that. (only 30 out of 330 remained)
I am doing a random effects regression now.

Since i have to explain why i didn't choose the fixed model i want to illustrate it. i thought that generating a variable a=1 if the person changed his/her mind during the survey waves and a=0 if the person didn't change his/ her mind.

i have a unbalanced panel so it is possible that a person only answered 3/5 surveys but nontheless a=0 if the ansers are the same.

Any ideas?

Thank you!
Best Katharina

[daniel klein]

See xtsum, which decomposes the variation of variables into between- and within-variation.

Best
Daniel

[Nick Cox]

Also an FAQ

http://www.stata.com/support/faqs/da...ions-in-group/

[Carlo Lazzaro]

As far as the theoretical underpins are concerned, Katharina may also want to take a look at Paul Allison's textbook on this topic http://www.stata.com/bookstore/fixed...els/index.html.