1

I have some data that- in its raw form - represents group binomial data.Y vector is the probability of an event. Using logistic regression I get one set of parameter coefficients.

Turning the data into a longer form - Bernoulli format (i.e. Y vector is 1 or 0), I use logistic regression again and get a different set of parameter coefficients from before, why is this?

Adam
  • 35
  • 4
  • 6
    See [here](http://stats.stackexchange.com/questions/144121/logistic-regression-bernoulli-vs-binomial-response-variables/144174#144174) where I've shown that the maximum likelihood estimates are the same (but deviance residuals and AIC will be different). When you input the Y vector as probabilities, you need to provide the number in each group as the weight (the way to do this differs between software). If you've done that, I suggest you check that you've converted the data to Bernoulli correctly. – Mark Mar 09 '16 at 21:15

0 Answers0