I'm dealing with some clustered data (clinical data on patients undergoing a specific procedure at several medical sites), and I need to account for a random effect through the site of intervention.
Given the fact I'm analysing both continous and binary categorical outcomes, i selected linear and logistic mixed-effect models as my models of choice, importing my covariate at fixed-effect level and including a random effect parameter for my clustering label.
Here comes the problem: I run this analysis in conda with Python 3.8 and, as far as I can see, statsmodel does support LMM (So i'm fine with my continous outcomes) but not binominal models. The only option available would be "BinomialBayesMixedGLM", but I'd rather stay within probabilistic if possible.
I tried using rpy2 package to access R packages within my python environment, but due to some incompatibilities I cannot solve with my current machine (I need to stick to MacOS11.7, which is dragging some more constraints in package updates), it doesn't work properly.
Any other approach for working with binary outcomes with a probabilistic mixed-effect model in python?
You can try out PyMC (or the more modern PyMC3/PyMC4). This is a Python library for probabilistic programming that can fit Bayesian models, including mixed-effects models for binary outcomes. Although it's Bayesian rather than frequentist, it is a strong alternative for mixed-effects modeling.
Thank you for your response. I’m currently updating my code in that direction, but i was still looking for a probabilistic option for dealing eith a binary outcome and a random effect. Thanks
You may consider using the R interpreter directly for this analysis. It will be easier than calling an R script from Python.
Here is an example R script:
if (!require(lme4)) install.packages("lme4")
data
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress