Need predictor to act as a factor

Asked Jun 10 '15 at 04:50

Active Jun 10 '15 at 04:52

Viewed 126 times

Trying to get the predictor "geology" to act as a factor...it says it is at the raster level, in the rasterstack, and in the dataframe, so not sure why it won't work.

is.factor(geology)
# [1] TRUE
is.factor(tstack$geology)
# [1] TRUE
is.factor(mbdata$geology)
# [1] TRUE

library(gam)
mel.gam.forw <- gam(mel.bur ~ s(aus35.s,3) + s(firefreq.s,4) + geology + 
                    s(slope.s,2) + s(wetness.s,3), family = binomial (link="logit"), 
                    data = mbdata)

mb.gam.f.pred <- predict(tstack, mel.gam.forw, type="response")

Error in `contrasts<-`(`*tmp*`, value = contrasts.arg[[nn]]) : 
  contrasts apply only to factors
In addition: Warning message:
In model.frame.default(Terms, newdata, na.action = na.action, xlev = object$xlevels) :
  variable 'geology' is not a factor

Any help would be much appreciated.

edited Jun 10 '15 at 04:52

asked Jun 10 '15 at 04:50

user4889553

My guess that is that `geology` is being coerced to something other than a factor. Could you check the other inputs to make sure that everything which should be a factor in fact is a factor? – Tim Biegeleisen Jun 10 '15 at 05:03
Geology is the only input that needs to be a factor – user4889553 Jun 10 '15 at 05:04
So it's failing on the call to `predict()` and not the call to `gam()` ? – Tim Biegeleisen Jun 10 '15 at 05:05
Please check to make sure that you don't have any `NA` values in the factor variable, q.v. [this SO post](http://stackoverflow.com/questions/20645926/contrasts-can-be-applied-only-to-factor) which discusses this. – Tim Biegeleisen Jun 10 '15 at 05:07
Saw that post before, tried it (geology <- na.omit(geology)) and checked it was still saying it was a factor...ran predict() again and same error @TimBiegeleisen – user4889553 Jun 10 '15 at 05:13
Isn't `predict` taking first "model", then "newdata" argument? – Roman Luštrik Jun 10 '15 at 05:14
is your predict statement correct ... notmall it is `predict(gam.object, gam.newdata ... `. You seem to have it reversed – user20650 Jun 10 '15 at 05:15
1

@user20650 that was a photofinish, well played. :) – Roman Luštrik Jun 10 '15 at 05:15
I'm not quite sure what you mean, sorry @RomanLuštrik – user4889553 Jun 10 '15 at 05:18
ha, and to you... now i need to do is work on my spelling – user20650 Jun 10 '15 at 05:18
try `predict(mel.gam.forw, tstack, type="response")` – user20650 Jun 10 '15 at 05:19
> mb.gam.f.pred <- predict(mel.gam.forw, tstack, type="response") Error in as.data.frame.default(data) : cannot coerce class "structure("RasterStack", package = "raster")" to a data.frame @user20650 – user4889553 Jun 10 '15 at 05:20
1

It would appear @user20650 and I were wrong. `predict` for `Raster*` is different (grrrr). See the [docs](http://www.inside-r.org/packages/cran/raster/docs/predict). – Roman Luštrik Jun 10 '15 at 06:43
Thanks anyway. Any other thoughts? Still stuck... @RomanLuštrik – user4889553 Jun 10 '15 at 06:47
1

At this point I will ask for a reproducible example. – Roman Luštrik Jun 10 '15 at 06:50

Need predictor to act as a factor

0 Answers0