Random slopes specification

A reviewer wants me to add random slopes for some of the fixed effects relative to another random effect. When I ran a model where the same random effect is correlated with two others I get an error. The model specification is below. I get the error:

“Duplicated group-level effects are not allowed. Occurred for effect ‘Intercept’ of group 'Verb”

I tried using lme4 instead and I dont get the error (though the model doesn’t converge). Do I need to specify the random effect structure differently? Am I writing the formula wrongly? Thanks

model1= brm(Case~ Affirmation+Affectedness+ Causative+ Telicity+ Tense+ Punctuality+
Person+ NumberObj+ Mood+ Kinesis+ NumberSubj+ AgencySubjAnimacyObj+ParticipantsCausative+
ParticipantsAgencySubj+AgencySubjCausative+ ConcretenessParticipants+ CountParticipants+
Group*Participants+
(1+Participants|Verb)+ (1+AnimacyObj|Verb)+ (1|Address),
data= training, family=bernoulli(link=“logit”),iter=3000,
cores=4, seed=123, save_all_pars = TRUE,
control = list(max_treedepth = 15), prior=set_prior(“cauchy(0,2.5)”, class=“b”)+
set_prior(“cauchy(0,10)”, class=“Intercept”))

Hi,

If I am not misunderstanding your question, then I think (1+Participants|Verb)+ (1+AnimacyObj|Verb)+ (1|Address) should be (1+AnimacyObj + Participants | Verb)+ (1|Address).

Moreover, I wondered what is the grouping variable (i.e. the unit within which the observations are clustered) here? For example, if you have a within-subject design where each participant provides multiple observations, you might want to add Participants as a random intercept so that (1 | Participants) would be part of your model. That however, depends on the context of your study.

It could also be the case that you have a study in which you measure a reaction of multiple participants to multiple verbs (and Address), than the observations would be cross-classified within participants and verbs so that you would have (1+AnimacyObj | Verb)+ (1 | Participants) + (1|Address).

This syntax in turn assumes that the variable AnimacyObj varies across Verbs but not participants. If it also varies across participants you would instead write (1+AnimacyObj | Verb)+ (1 | Participants) + (1|Address)

I hope this answers your question, but maybe you also know all of this already and I misunderstood your question, so if this is the case let me know :)

Hey, thanks for your reply. This is not experimental data, it is corpus data (5000 sentences) so the variable Participants does not refer to subjects in an experiment but whether a verb is Transitive (2 participants) or Intransitive (1 participant). I’m using a certain theory that uses that term in that way. I’m looking at causative constructions such as “she let me use her book” and “she made me use her book”. So the grouping variable here is Verb (the embedded verb in each data point/ sentence) as the same verb can appear in multiple observations (I let her eat the pizza; I made them eat the vegetables; She let me eat some of her dessert; etc).

So with this new information, the correct syntax is still: (1+AnimacyObj+Participants|Verb)+ (1+Address)?

Thanks again!

Ah okay, I see, sorry for confusion. Yes, I think it should be (1+AnimacyObj+Participants|Verb) + (1|Address) then.

No problem, the name of the variable is very misleading! Thank you for your help and quick reply