Problem with mixed effects model with binary data

pydbl

New Member
#1
Hi guys,

I have a question about logit mixed effects model. I tried to use lmer to fit a model with binary dependent variable (Yes or No), a categorical fixed predictor with 3 levels, and two random effects. However, it turned out that one level of the fixed resulted in only positive outcomes (so all Yes's). As a result, the result obtained from the model I fit with lmer indicates that there is no significant difference between the level with all positive outcomes and the other levels. But this is obviously not true since the positive outcomes for the other two levels were only 20% and 80% respectively. So I wonder if you guys know how this problem may be resolved in R. For example, what other alternative approaches can be taken. Thank you in advance for your help!
 

Dason

Ambassador to the humans
#2
What are the sample sizes? Just because one level gives you all positive responses doesn't mean you can say there is a statistically significant difference automatically.
 
#3
What are the sample sizes? Just because one level gives you all positive responses doesn't mean you can say there is a statistically significant difference automatically.
I had 16 participants, and each of the 3 levels of the fixed predictor had 8 items.