[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: issues about the consistency of parameter estimation

From   Maarten buis <>
Subject   Re: st: issues about the consistency of parameter estimation
Date   Thu, 29 Oct 2009 06:12:19 -0700 (PDT)

--- On Thu, 29/10/09, amatoallah ouchen wrote:
> I'm tying to estimate  4 parametres a1, a2, beta ,
> and sigma, I'm  expecting that beta values should
> be postive and significantly different from 0,
> sometimes it's postive but Not statistically
> significantly different from zero , on other cases it's
> negative and Not statistically significantly different
> from zero.

That depends on the nature of your expectation. For 
instance we know that a variance, by definition, cannot be 
negative, so if we find a program that results in negative
estimates of variances, we know that the program must 
contain a bug. In other cases we may, based on our substantive
theory, have a hypothesis that the coeficient should be
positive. In that case finding negative or non-significant
results is be just an empirical finding indicating our 
theory is wrong. Another interpretation of non-significant
results are that you just don't have enough observations
to find what you are looking for.

We cannot tell you that your results are ok, or that you
are suffering from "problem xyz". Making any such diagnosis
would require us to know exactly what your data looks, 
what you did to it, what the general context of your research
is, etc. etc. We can give you some general advise on where
to look and what to look for, as I have just done, but in
the end you are the one that will have to make the decison
whether or not your results are OK. (Than it is up the 
pear-review process to find all the errors you made.)

-- Maarten

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen

Send instant messages to your online friends 

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index