News for hackers. Stuff that matters.
Re:The Sigmoid Function in Logistic Regression by nabla9 on Tuesday May 17, 2016 @07:28AM (#11712666)
y = a0 + Σ(ai×xi)

Use linear regression when 1 unit increase in x increases y by a.

Use logistic regression when 1 unit increases x increases log odds by a.

Use Poisson regression when 1 unit increase in x increases y by y×e^a (see also negative binomial regression)

Re:The Sigmoid Function in Logistic Regression by nimish on Tuesday May 17, 2016 @08:05AM (#11712862)
Important that these are on average.

https://en.wikipedia.org/wiki/Generalized_linear_model

Re:The Sigmoid Function in Logistic Regression by gbrown on Tuesday May 17, 2016 @10:07AM (#11713728)
Binomial and Poisson regression can both be performed with non-canonical link functions, which sometimes provide a better fit or more interpretable results (see: canonical link for gamma regression).