r - Negative binomial modelling error -
i have been asked lecturer fit negative binomial model data (looking @ distance patch of forest moth observations occurred @ different status groups (endemic, exotic, native).
my dataframe looked (n=2466):
id native.exotic urban_non bush.distance scientific_name 12469 exotic 0 0 opodiphthera eucalypti 224769 native 1 170 agrotis ipsilon 251928 native 1 405 platyptilia falcatalis 524208 exotic 1 2010 ctenoplusia limbirena
i advised use mass , run code
m1 <- glm.nb(count ~ bush.distance*native.exotic, data=moth) drop1(m1, test="chisq") xv <- (with(moth, seq(from=min(bush.distance), max(bush.distance), length.out=200))) newdat <- expand.grid(bush.distance=xv, native.exotic=levels(moth$native.exotic)) newdat$fit <- predict(m1, newdata = newdat, type="response") + geom_line(data = newdat, aes(x=bush.distance, y=fit))
however, have no count variable. used melt convert data long format looks this:
dist status count 0 endemic 844 1 endemic 8 5 endemic 3 10 endemic 5
but still when run code, result:
m1 <- glm.nb(count ~ dist*status, data=bush_status) error: no valid set of coefficients has been found: please supply starting values
my knowledge of r super basic apologise if haven't explained appreciated. thanks!
Comments
Post a Comment