I'm trying to run a Gamma analysis in a self-paced reading data. However, the model successively fails to converge. I've seen some answers here trying to solve this problem for other people, but none of the solutions was feasible for me.
I couldn't get an example from my data because it's too long (more than 9k rows). It consists of a self-paced reading experiment in which TR is the response time and I want a gamma model (since the data is not normal at all, but skewed to the right) to calculate TR in function of cond (condition; there were 3 in the experiment (C1, C2 and C3)) and occurrence (there were also 3: OC1, OC2 and OC3).
I've already tried to use a Nelder_Mead optimizer instead of the default optimizer, but it seems not to work either.
Take a look at my models:
mod_cond*oc <- glmer(TR ~ cond*oc + (1|item) + (1|sujeito) , data=dadoslimpos5, family=Gamma(link="identity"))
mod_cond*oc <- glmer(TR ~ cond*oc + (1|item) + (1|sujeito) , data=dadoslimpos5, family=Gamma(link="identity"), control = glmerControl(optimizer ="Nelder_Mead"))
None of them seems to work. I've gotten these messages for each of them, respectively:
Warning message: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model failed to converge with max|grad| = 0.0484605 (tol = 0.002, component 1)
Warning message: In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv, : Model failed to converge with max|grad| = 0.0471255 (tol = 0.002, component 1)
Is there any solution to this problem?
If someone needs my data to take a look at, I've posted the clean data (only regions of interest) here. I've already published these data before using mixed models, but now I need to take a look at them using gamma and I'm expecting the model to converge without warnings.
Thanks a lot!