Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: Comparing Regression Coefficients Across Subsamples

From   David Ashcraft <>
To   "" <>
Subject   Re: st: Comparing Regression Coefficients Across Subsamples
Date   Mon, 6 Aug 2012 07:08:49 -0700 (PDT)

Thanks Nick,

Aside from regional affiliation, firms should be very similar as each firm is an investment firm with single mandate. I am using  _xtreg for regression analysis on the overall firms and then the same on each individual firm. R2 is 0.90 for almost every firm in the sample. Now I want to know if there is any difference in coefficients based on the regional affiliation. Based on  this situation what is your recommendation?

----- Original Message -----
From: Nick Cox <>
Sent: Monday, August 6, 2012 2:59:29 PM
Subject: Re: st: Comparing Regression Coefficients Across Subsamples

Several questions are bundled together here. Working backwards from
the easiest:

3. Analysis of variance is the method centred on the comparison of
means of several groups.

2. However, your dataset is a panel dataset gathered over time and so
plain analysis of variance is unlkely to be suitable. Much depends on
quite what dependence structure you think underlies your data and thus
your regressions (of what on what?).

1. I can't comment on Cohen's recipes as I have not read the paper cited here.


On Mon, Aug 6, 2012 at 12:07 PM, David Ashcraft
<> wrote:

> I am working on a panel data of more than 200 firms with distinct "id" over 20 years with monthly frequency. Each firm belongs to one of the three groups (1,2 and 3) based on regional affiliation. Exit and entry is unrestricted. The resulting data set in unbalanced. I Now I want to compare the regression coefficients of each regional group in the light of discussion provided in:
> Cohen,
> A. (1983). Comparing Regression Coefficients Across Subsamples A Study of
> the Statistical Test. Sociological Methods & Research,vol. 12(1) pp 77-94.
> doi: 10.1177/0049124183012001003
> I believe -ttest- i.e. differences in means analysis over coefficients of each group is what Cohen A . (1983) is referring to. Am I right about this? Please tell me, if there is any other command that can be used to implement Cohen A. (1983)
> Aside from that _ttest- compare means between two groups at a time. In my case what should I do to test for three groups at the same time?

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index