S Roche Hendrix
1st March 2005, 04:58 PM
The question below came to me from a chemist. I did not have an answer for him but before researching would like to see if any of you do.
"I was performing regression analysis in excel and minitab for x & y variables. I had 24 trials and under Data Analysis/Regression in excel the reported degrees of freedom was 21 in the summary table. In minitab the same regression analysis reported 22 degrees of freedom which makes sense to account for the two variables. Why the difference? In reality it doesn’t make much difference with 24 trials, but if the trial number was low then the difference would be greater."
S
Al Dyer
1st March 2005, 05:12 PM
Possibly rounding differences in the programs???? I would trust Mini-Tab over Excel on general purposes ans excel was not specifically designed for statistical use.
And as we all know, it was designed by microsoft.:lol:
Al...
Steve Prevette
1st March 2005, 05:56 PM
Do you have a copy of the Excel file that you could post here?
S Roche Hendrix
2nd March 2005, 09:56 AM
Steve,
I asked for the data and received the reply below. Thanks anyway. I appreciate your willingness to help.
Susan
"I did not save all of my regression data only the summary table. I went in to regenerate the entire regression analysis for the data, and guess what? Each time I do it now I get the correct degrees of freedom. I do not understand. I did this multiple times before and it was reporting the same number of observations as now “24”, but only 21 df. Now its reporting 22 df??? Sorry for bothering you, I don’t know what has changed but everything is fine now."
Steve Prevette
2nd March 2005, 10:53 AM
I do not understand. I did this multiple times before and it was reporting the same number of observations as now “24”, but only 21 df. Now its reporting 22 df??? Sorry for bothering you, I don’t know what has changed but everything is fine now."
It is possible that they had inadvertently done a non-linear regression (a parabola) or otherwise got excel to believe there was a second variable which would have used up the 3rd degree of freedom.
QualityJedi
2nd March 2005, 01:50 PM
I have had similar experiences with MS Excel. It is possible to write a few lines of VBA Code to make the numbers (correctly) match MiniTab, however, it is a pain in the keester. The root cause of the discrepancy usually lies in the number of significant digits allowed in the calculation and the rounding of same.
MiniTab is the undisputed champion and de-facto standard in todays business world with a price tag of $1,000+ depending on licensing.
Mike H.
Quality Jedi
Sales...Quality
:biglaugh: :whip:
