Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: st: RE: variance in glm
From
Nick Cox <[email protected]>
To
"'[email protected]'" <[email protected]>
Subject
RE: st: RE: variance in glm
Date
Wed, 2 Mar 2011 17:53:33 +0000
-egen- has a -sd()- function. So in principle you just need to square to get variance. But I don't think this helps you much in practice, as it can't give you results for your singleton groups.
On the other hand, variations in variance don't seem to bite as much as many people guess they might. A way to explore this is to model using differing variance assumptions and see how much difference it makes to your substantive conclusions. Even though there might be no obvious model for your situation, going non-parametric on this just sounds like a cul-de-sac.
Nick
[email protected]
Argyn Kuketayev
let me reformulate the problem then.
i have a set of observations (Xi, Yi). Xi and Yi are both discrete
numbers. my current hypothesis is that Yi = b1 Xi + b2 + Ei, where Ei
is an error term, maybe Gaussian. I think that Var[Ei] is not
constant, and want to estimate it somehow. the sample size is ~500,
and some Xi values repeat in observations, up to 30-40 times, i.e. i
could compute Var[Yi] for some Xi values directly, on the other hand
many Xi values don't repeat, i.e. there's only one observation with a
given Xi.
so, how to compute Var[Ei] for a any given Xi in Stata? preferably in
some non-parametric way
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/