Date: Dec 12, 2012 11:06 AM
Author: Bruce Weaver
Subject: Re: Multiple regression with all dummy variables
On 11/12/2012 7:10 PM, Bruce Weaver wrote:
> On 11/12/2012 5:22 PM, Gary wrote:
>> On Tuesday, 11 December 2012 20:20:48 UTC+2, paul wrote:
>>> Does a multiple regression with all dummy (indicator) variables make
>>> sense? I work at a state university tutoring various basic subjects
> --- snip ---
>>> Thanks for any help!
>> I think you can find some of the argument in
>> Cohen, J. (1968). Multiple regression as a general data-analytic
>> system. Psychological Bulletin, 70, 426-443.
>> Also Cohen's famous textbook.
> See also Judd & McClelland's book "Data Analysis: A Model Comparison
> Approach" if you can find a copy.
Oops...I also meant to comment on this bit from the OP:
"But the students are then told that the multiple regression gives more
information since we can conclude from the t-tests on individual
coefficients that silver cars sell for more than the base case (black.)"
Most ANOVA programs have various methods for making pair-wise
comparisons, and many of them could make the same set of comparisons
captured by the t-tests in the table of regression coefficients. In the
case where each of k-1 treatments is compared to a control group, many
experimentalists would probably use Dunnett's test, which was designed
for that situation.
The point is that the ANOVA program *can* give just as much information
(and more) than the info captured by the regression coefficients.
"When all else fails, RTFM."