[R-sig-ME] Improving residual distribution in glmmTMB

[Timothy MacKenzie]

Greetings,

I'm modeling the number of downloads (downs) of certain forms
(Document) by a number of school districts (District) of various
student sizes (ELs).

I've tried genpois(), nbinom1(), and nbinom2() families but the fitted
vs. residual doesn't look promising.

In glmmTMB, are there any additional ways to improve the residual
distribution for count-type data?

Thank you,
Tim M
### Reproducible data and code:
d = read.csv("https://raw.githubusercontent.com/fpqq/w/main/b.csv")

model = glmmTMB(downs ~ I(ELs/20) + Document +
                           (1|District) +
                           (1|Document),
                         family=genpois(),
                     #  ziformula = ~ Document,
                  # dispformula = ~ Document,
                         data = d)

plot(resid(model,type="pearson") ~ fitted(model))