Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Topic: Discrepancy between t-test and regression analysis outcome?
Replies: 4   Last Post: May 22, 2006 7:01 PM

 Messages: [ Previous | Next ]
 Paige Miller Posts: 46 Registered: 3/18/05
Re: Discrepancy between t-test and regression analysis outcome?
Posted: May 21, 2006 11:56 AM

On 5/21/2006 5:48 AM, TLW wrote:
> I have been working on the data analysis of a group of data for two
> different (though related) problems.
> First we were interested in determining whether the data could be
> considered homogeneous.
> Secondly we were interested in predictions based on the data.
>
> So far I have assumed that those two problems should give 'similar'
> answers. I.e.: if a t-test shows that groups A and B do not differ;
> then a regression test using the data in group A should give (more or
> less) the same model as a regression test with group B.
>
> However, in my data this is not true.

<...snip...>

> * Am I right that the identification of 'no difference' between two
> positions should mean that those two positions should give
> (approximately) the same model using stepwise regression?

Without going into detail about what you did (cause I didn't
understand everything), when you do a t-test of differences, and the
result is no difference, that means that the MEAN of the two groups is
not significantly different. It says nothing about the regression
itself. The regressions can be dramatically different around the same
mean. You should be able to draw an example on a piece of paper...

You might want to fit a regression to each group and test to see if
the groups have the same intercept and test to see if the groups have
the same slope.

--
Paige Miller
pmiller5@rochester.rr.com

It's nothing until I call it -- Bill Klem, NL Umpire
If you get the choice to sit it out or dance,
I hope you dance -- Lee Ann Womack

Date Subject Author
5/21/06 Paige Miller
5/21/06 TLW
5/22/06 Paige Miller
5/21/06 Richard Ulrich