0

I have a response variable X with values 0 and 1. I can use it as an integer or as a factor in regression.

I use brms package for modelling. The families of brms are: https://rdrr.io/cran/brms/man/brmsfamily.html

How should I model the X variable? Using binomial or bernoulli family? Both are logistic regressions?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
st4co4
  • 1,499
  • 5
  • 10

1 Answers1

2

Bernoulli family is a special case of the binomial family, so it should not matter which you use (but maybe the model specification is simpler with Bernoulli family, as you do not need to specify $n$.) Both specify the same logistic regression model and should lead to identical results.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
  • Generally speaking, does it mean that they all are just logistic regressions? – st4co4 Oct 06 '20 at 17:23
  • 1
    Yes, just logistic regression. Sometimes you have individual-level data you can choose to do as a Bernoulli family, or you can group the data and choose a Binomial family. The regression model will be the same, but (at least in R) the results will be presented differently. For details https://stats.stackexchange.com/questions/27400/logistic-regression-grouped-and-ungrouped-variables-using-r – kjetil b halvorsen Oct 06 '20 at 17:34