[R] nnet Multinom output of ordered predictors
DCharbonneau
charbonneau.daniel at gmail.com
Sat Apr 23 20:02:45 CEST 2011
Hello,
I apologize if this seems like an obvious question, but I have been looking
everywhere and have yet to find an answer.
I am doing a multinomial regression with multinom() in the nnet package. I
have a 3 level ordered response (ordered()) variable and 4 predictors, 3 of
which are numerical and one which is an ordered factor (also ordered()) with
5 levels (a, b, c, d, e).
My question is in regards to understanding the coefficient names of the
ordered variable in the output. Specifically, I get this as the coefficient
output of the function
Coefficients:
(Intercept) ageNum.L ageNum.Q ageNum.C ageNum^4
couvert.paysage.1500 heterogen.couvert.paysage.1500 ac
1 -3.741361 0.3574398 -0.9131615 0.02510015 0.13612309
-0.3204320 2.244231 2.357034
2 -2.019076 1.2342684 -1.3330470 0.53676536 0.04868537
-1.3867078 1.322942 4.115914
I am simply wondering what the ageNum.L, ageNum.Q, ageNum.A and ageNum^4
refer to?
As it is an ordered variable, can I assume that multinom contrasts a to b, b
to c, c to d and d to e?
If I perform multinom with the same formula, but with the ordered predictor
forced as a factor, I obtain this:
Coefficients:
(Intercept) ageNumb ageNumc ageNumd ageNume couvert.paysage.1500
heterogen.couvert.paysage.1500 ac
1 -4.447215 0.7876454 1.291553 0.9819608 0.4679982 -0.3204237
2.244232 2.357034
2 -3.676090 1.9392234 2.404509 2.0408820 1.9006754 -1.3867094
1.322924 4.115888
in which case it seems obvious that levels b, c, d and e are contrasted to
level a.
I really appreciate any help I could get. Thanks for you time!
--
View this message in context: http://r.789695.n4.nabble.com/nnet-Multinom-output-of-ordered-predictors-tp3470423p3470423.html
Sent from the R help mailing list archive at Nabble.com.
More information about the R-help
mailing list