From: <tastard_at_cict.fr>

Date: Mon 14 Mar 2005 - 18:29:28 EST

mail2web - Check your email from the web at http://mail2web.com/ .

R-help@stat.math.ethz.ch mailing list

https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html Received on Mon Mar 14 18:36:56 2005

Date: Mon 14 Mar 2005 - 18:29:28 EST

Can you help me interpret the output I get with “polymars”?
I'd like to find a regression model to predict 2 dependent variables (then
called Y1 and Y2) with 2 independent variables (then called X1 and X2).

Here is the output:

polymars(responses = data[, 13:14], predictors = data[, 2:3])

Model fitting

0/1 size RSS 1 RSS 2 GCV 1 1 1 0.2486744 0.6520499 0.01756785 2 1 2 0.2316042 0.3820036 0.01391882 3 1 3 0.2312835 0.3819480 0.01637875 4 1 4 0.2299605 0.3766946 0.01935784 5 1 5 0.2243804 0.3766148 0.02331277 6 1 6 0.2243499 0.3764709 0.02893749 7 0 5 0.2243804 0.3766148 0.02331277 8 0 4 0.2299605 0.3766946 0.01935784 9 0 3 0.2312835 0.3819480 0.01637875 10 0 2 0.2316042 0.3820036 0.01391882 11 0 1 0.2486744 0.6520499 0.01756785

Model produced

pred1 knot1 pred2 knot2 Coefs 1 Coefs 2 SE 1 SE 2 1 0 NA 0 NA 0.40139070 0.14049628 0.02195431 0.02819553 2 2 NA 0 NA -0.02146696 -0.08538267 0.01047337 0.01345076

**RESPONSES : 2
**
Rsquared : 0.069 0.414

If I have well understood, polymars tries to build several models for each dependent variable (with one function if it is homogeneous or more ones if it is not homogeneous) and selects the best one. I suppose the model is the best one when the GCV is smallest. So here I thought I should have had a model with 2 functions for each dependent variable (size 2).

Why do I get only one result (1 function for each dependent variable)? How can I get a model with different functions over different ranges of the independent variables? Do you think it would be a better model?

Are these the good equations?

Y1 = 0,40139 – 0,02147 X2

Y2 = 0,1405 – 0,08538 X2

Does polymars take the interaction between Y1 and Y2 into account ?

Thanks

mail2web - Check your email from the web at http://mail2web.com/ .

R-help@stat.math.ethz.ch mailing list

https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html Received on Mon Mar 14 18:36:56 2005

*
This archive was generated by hypermail 2.1.8
: Fri 03 Mar 2006 - 03:30:47 EST
*