Today , I get the terrible resulits from the glmer.nb(). I tried a plent of methods to change the variables. However, When I finaly fitted a model with FUll variable . qqPlot(resid(Model)) tell me I got the right things but DHARMa's result told me I'm wrong. Shall I believe DHARMa Pakacages? Of course , I should appreciate the author for his contribution to DHARMa which is plenty easier to validate model, but today's result makes me confused.
These are the prodecures that I cope with my dataset:
1.rm(M) #first I want to remove the case in case of unneccesary troublesome.
2.M = glmer.nb(z ~ x + y + RC1 + RC2 + (1 | SITE) +
(1 | PID) + (1 | YEAR) + (1 | DATE), data = dt, verbose=TRUE,glmerControl(optimizer = c('bobyqa')))
3.simulateResiduals(M,n = 1000) %>% plot()
and DHARMa tells me how terrible things are, like this!:
4.but if I use qqPlot , it tells me like this, I think my model is ok, right? But DHARMa told me not. I don't know how to believe . Anybody could make a good interprete?
par(mfrow = c(1,2))
qqPlot(resid(M))
plot(resid(M) ~ fitted(M)); abline(h=0)
And pardon me I can't upload my dataset for your convinience because it will exceed the number of character. If you know how to deal with it , please tell me . Thank you very much.
