Dear all,
I want to model presence/absence data of tree occurrence using a number of
predictor variables.
Absences of some sample points are probably wrongly specified (they should
be presences) due to land use which can not be incorporated as a predictor
because of some sort of arbitrariness. Some trees were logged while others
were not but to both cases land use as a category would apply. However, at
last I cannot decide if absences occur truly because of unsuitability of
conditions at sampling location. Therefore, I want to downweight these
"wrong" absences based on a certain algorithm producing decimal
numbers so
that they become less influential (i.d. I have a column "weights"
(1>=weights>0) of the same lenght as predictors and response variable). I
tried to figure this out checking the forum, but are not sure about it.
I guess, the weight argument in glm-function does not do what I intend to
do? Might the survey package be a solution? And if so, do I ignore all the
arguments except weight to specify the svydesign?
Thanks a lot for your help!
Eva
--
View this message in context:
http://www.nabble.com/Downweighting-of-cases-in-GLM-tp16823770p16823770.html
Sent from the R help mailing list archive at Nabble.com.
