3 ms·
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.
by nabla9 11y ago
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)
- nimish 11y agoImportant that these are on average. https://en.wikipedia.org/wiki/Generalized_linear_model https://en.wikipedia.org/wiki/Generalized_linear_model
- gbrown 11y agoBinomial 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).