I'm curious what typical software packages do if the covariate is continuous?

Essentially you need to maximize

$logit(P(Y=1 | X)) = \alpha + \beta*x$

for the simplest case.

However, but if X is continuous how exactly do we calcualte the probability to put into the logistic function since we don't know it?

Once we have it, it's easy to minimize $\sum_i (logit(Pr(Y_i = 1| X)) - \alpha + \beta * x_i)^2$ but I'm unclear how the machine mechanically calculates the probability?

Thank you!


You don't estimate $\beta$ by least squares in logistic regression. You estimate $\beta$ by maximizing the bernoulli likelihood.

| cite | improve this answer | |
  • $\begingroup$ Argh -- you're right. And I knew that which makes it even more frustrating. Thanks for the reminder! $\endgroup$ – user1357015 Apr 22 '16 at 18:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.