Chronological Thread 

< Chronological >      < Thread >

---

Hi Laura,

Yes, every covariate in the GLM formula must be provided during the
prediction step, either as a raster or as a constant value. If you omit a
covariate that was included in the model formula, the prediction tool will
fail. If you have terms that are not very important for the model or that
should be removed (e.g. the p value is > 0.05) then you should rerun the Fit
GLM tool without those terms and produce a model with just the terms that you
really care about.

For many projects, all of the terms are geographically-varying and should
therefore be provided as rasters (e.g. bathymetry or elevation). Terms that
do not vary can be provided as constant values. It sounds like this might be
what your total size parameter might be. The value you pick depends on what
question you're asking. In some situations you might just provide an average
value. In other cases, maybe you want to run multiple predictions, each for
separate values (e.g. small, medium, and large sizes), and see how they
compare.

Best,

Jason

-----Original Message-----
From:


[mailto:]
On Behalf Of

Sent: Saturday, May 14, 2016 4:33 PM
To:

Subject: [mget-help] Predict GLM from rasters

Hi,
I was curious if it was necessary to have all the variables from the GLM
formula input into the "Predict GLM from rasters" tool. For instance, one of
my predictor variables was total size (in cm, continuous). Can it be used in
the tool, or would I have to exclude it? And if I excluded it, would that
alter the overall raster output?

Thank you,
Laurel
To unsubscribe from this list, visit:
https://lists.nicholas.duke.edu/sympa/ca/initiate_unsubscribe/mget-help/jason.roberts%40duke.edu

---

• [mget-help] Predict GLM from rasters, llam, 05/14/2016
  • RE: [mget-help] Predict GLM from rasters, Jason Roberts, 05/15/2016

---

Archive powered by MHonArc 2.6.19.