[R-sig-eco] ordinal scale data (not respons variable) in linear regression
Herczeg Róbert
robert.herczeg at gmail.com
Wed Jan 30 09:19:28 CET 2013
Dear members,
I have a question about how should I handle ordinal scale data (when
it is not the respons variable) in linear regression or GLM,GAM? (R
handles ordinal data as a factor)
Should I use as a factor? In that case I lose information and it is
not appropiate especially if exits replicates.
If I use in interval scale (in a given distance between the ordinal
data) I think, it is also not correct.
For example:
Respons variable is count data (for small mammal) and background
variable is (of course not this only one, but this is in ordinal
scale):
- vegetation cover (<25% few, 25%<middle<50%, 50%<high<75%, 75%<total)
As a fator (4 level):
few
middle
high
total
In interval scale:
few = 1
middle = 2
high = 3
total = 4
Sorry for my english, it is not my mother language.
Kind regards,
Róbert
--
Róbert Herczeg PhD student
University of Pécs
Faculty of Sciences
Deparment of Animal Ecology
Ifjúság útja 6., Pécs H-7624, Hungary
(http://www.ttk.pte.hu/biologia/zootax/)
More information about the R-sig-ecology
mailing list