Polymars questions

["Ho-Won Jung" <hwjung@korea.ac.kr>]

Dear S-Plusers;

 

I am a S-Plus 2000 user, Windows XP

 

Would you do me favor for interpreting two outputs (classify=F, classify=T) of polymars?

 

I got output from MARS (polymars library in S-Plus)with dataset copied from

http://lib.stat.cmu/edu. The dataset in the web site is different from that used

in a manual included in polymars library for S-Plus.

 

Using the dataset that I got, a S-Plus program and results of polymars are as follows:

 

QUESTION 1 ---------------- Program --------------

> index<-sample(nrow(boston.dat),nrow(boston.dat)/2)
> dni<-matrix(c(5,6,3,11),byrow=T,ncol=2)

 

> boston.mars<-polymars(boston.dat[index,14], boston.dat[index,-14],
+ knots=15,factor=4, ts.resp=boston.dat[-index,14],
+ ts.pred=boston.dat[-index,-14], no.interact=dni)

------------ Outputs--------

Model produced

 

   pred1 knot1 pred2 knot2     coefs      SE
 1     0    NA     0    NA -41.13515 20.7118
 2    13    NA     0    NA  -1.29188  0.3118
 3    13  7.12     0    NA   1.75726  0.3229
 4    11    NA     0    NA   1.89812  0.8410
 5     7    NA     0    NA   0.61651  0.1160
 6     6    NA     0    NA  12.91582  3.1466
 7     6  6.49     0    NA   6.24186  1.2262
 8     1    NA     0    NA  -0.10910  0.0253
 9     7    NA    13    NA  -0.01521  0.0026
10     6    NA     7    NA  -0.07088  0.0157
11    13 15.55     0    NA   0.53258  0.1582
12     6    NA    11    NA  -0.39358  0.1333

 

Rsquared : 0.88
>

 

-------- Interpretation------

 

Is the following correct interpretation of the above output?

Please check it.

 

 

Response = -41.13 - 1.29 x13 + 1.75 (x13-7.12)+  1.89 x11 + 0.61 x7 + 12.91 x6 + 6.24 (x6 - 6.49)+ - 0.11 x1
           -0.01 x7 * x13 - 0.07 x6 * x7 + 0.53 (x13 - 15.55)+ - 0.39 x6 * x11

 

------- Rewriting -------------

 

Response = -41.13      - 0.11 x1  + 12.91 x6 + 6.24 (x6 - 6.49)+ + 0.61 x7 + 1.89 x11
           - 1.29 x13  + 1.75 (x13-7.12)+  + 0.53 (x13 - 15.55)+
    - 0.07 x6 * x7  - 0.39 x6 * x11  -0.01 x7 * x13

 

 

QUESTION 2 ----------

 

The following example is from a polymars example embedded.

 

 

state.pm<-polymars(state.region,state.x77,knots=15,classify=T)

 

 

pred1 knot1 pred2 knot2  Coefs.1    Coefs.2  Coefs.3  Coefs.4   SE.1   SE.2   SE.3   SE 4
1     0    NA     0    NA  3.47652  0.0418566 -0.96302 -1.55535 0.4588 0.4109 0.5735 0.6064
2     6    NA     0    NA -0.06191 -0.0009963  0.02456  0.03835 0.0095 0.0085 0.0118 0.0125
3     6  53.3     0    NA  0.04230  0.0870996 -0.0

 

Would you let me know how I can interpret the following?

 

 

 

 

If you know any documentation that covers MARS, would you let me know.

 

I really appreciate you.

 

 

 

Best regards,

 

Ho-Won Jung.