[R-sig-ME] Fitting a zero inflated mixed-effects mode to proportion data using glmmTMB
Sonia Bejarano
@on|@@bej@r@no @end|ng |rom |e|bn|z-zmt@de
Fri May 3 17:13:46 CEST 2019
Dear R-sig-mixed-models Listers
I am trying to fit a zero inflated mixed-effects model to determine
whether my response variable (i.e. proportion of organisms with
parasite) differs between 2 conditions (protected zones vs. open zones),
and between 2 zones (inner, outer). Aside from these two fixed factors,
I have 2 random factors (i.e. transect and site). As my response
variable is zero-inflated (i.e. I have 41% observations are 0) I tried
fitting:
MZINF<- glmmTMB(organisms.with.parasites/total.num.organisms.checked ~
condition + zone + +(1 | site)+ (1 | transect), + zi=~., + disp=~zone, +
family = binomial, + weights=total.num.organisms.checked, + data=zanvic)
Here my questions:
1) if I fitzi = ~. or evenzi = ~ condition + zone I get convergence problems and NAs
but I get no NAs or convergence problems when I fitzi = ~ 1, zi = ~ condition, orzi = ~ zone
does this mean that due to insufficient observations per condition or per zone, I cannot test
simultaneously whether the probability of observing a structural 0 is influenced by zone and by condition?
2) If when I fitzi = ~ 1
Zero-inflation model:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -21.37 12320.75 -0.002 0.999
Does the p value > 0.05 values in here mean that the probability of observing structural zeroes
is independent of by zone or condition? or is*not *independent of zone or condition?
2) Different from the outputs explained in Brooks et al 2019 (The R
Journal Vol. 9/2, December 2017) when fitting the above I do not get a
Dispersion model section in my output, why is this? have I coded
something wrong in my model?
Many thanks,
Sonia
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list