Re: MLE
- From: hrubin@xxxxxxxxxxxxxxxxxxxx (Herman Rubin)
- Date: 30 May 2008 17:00:16 -0400
In article <2b54319b-f211-46bd-b8ad-39c16dd548be@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
C Hanck <chrhanck@xxxxxxx> wrote:
I want to know in case of OLS we made strong assumptionare you saying you need stronger assumptions for OLS to work? that is
about the distribution of error term. Why not in MLE? Is there any
assumption we are making in MLE which is of same type as of error
assumption in OLS?
clearly not correct. you may need something like E(u^2)<infty for OLS
whereas you assume normality of u for MLE, which implies that you
assume all moments of u to exist.
The estimates under OLS have the same asymptotic properties
whether or not the errors are normal, as long as they have
reasonable distributions (finite variance) and are uncorrelated
with the explanatory variables. This is a robustness property
of OLS, which is MLE under normality.
However, the condition that the errors be uncorrelated with
the variables is important. In structural estimation, it is
usually the case that they are not.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
hrubin@xxxxxxxxxxxxxxx Phone: (765)494-6054 FAX: (765)494-0558
.
- Prev by Date: Re: The danger of classical hypothesis and significance tests [was Re: MADLY AMUSED]
- Next by Date: Re: Levy process questions
- Previous by thread: Re: MLE
- Next by thread: Mean/ Median
- Index(es):
Relevant Pages
|
