Excel & MINITAB generated regression equations are different
 

I'm analysing a power series (y = a*x exp(b)). After analysing the data in
Excel, I re-ran the data in MINITAB. The b's are the same, but the a's vary
by a factor of 20! I need to reconcile these two results. any suggestions?

Earl Eiland

Earl -

I'm analysing a power series (y = a*x exp(b)). After analysing the data
in Excel, I re-ran the data in MINITAB. The b's are the same, but the a's
vary by a factor of 20! I need to reconcile these two results. any
suggestions? <

Possible explanation: Excel's trendline feature (you didn't specify how you
analyzed the data in Excel) fits the power model by first taking natural log
transforms of the x and y data. Then it finds coefficients that minimize sum
of squared deviations between actual Ln(y) and fitted Ln(y). Perhaps Minitab
searches for the coefficients that minimize sum of squared deviations
between actual y and fitted y; you could do this in Excel using Solver.

Suggestions: (1) Be sure to chart the data and look at the pattern to verify
that the power model is appropriate. (2) Don't use a Line chart type if
you're using Excel's trendline feature for the fit; the Line chart uses x
values of 1,2,3,... for the fit; it doesn't use the actual x labels;
instead, use an XY (Scatter) chart, in which case you may find better
agreement between Excel and Minitab.

- Mike
www.mikemiddleton.com

I would be happy to look at the data - send file to my email NOT to the
newsgroup
--
Bernard V Liengme
www.stfx.ca/people/bliengme
remove caps from email

"Earl E" <Earl wrote in message
...
I'm analysing a power series (y = a*x exp(b)). After analysing the data
in
Excel, I re-ran the data in MINITAB. The b's are the same, but the a's
vary
by a factor of 20! I need to reconcile these two results. any
suggestions?

Earl Eiland