\( f(x|y) \)- conditional distribution x gave as y which is 0 or 1. | Suggested model terms |
---|---|
The standard, common variance i.e. \( Var(x_{j} |y = 0) = Var(x_{j} |y = 1) \) | \( X_{j} \), i.e. the predictor itself These values imply that i \( X_{j} \) t is NOT customarily distributed we might consider transforming \( X_{j} \) to approx. Normality. |
Normal, unequal variances i.e. \( Var(x_{j} |y = 0) \ne Var(x_{j} |y = 1) \) | \( X_{j} \) and \( X_{j}^{2} \) |
Skewed right | \( X_{j} \) and \( \log_{2} \left( {X_{j} } \right) \) Log base 2 is more comfortable to interpret |
\( x \in [0,1] \) | \( \log_{2} \left( {X_{j} } \right) \) and \( \log_{2} \left( {1 - X_{j} } \right) \) |
\( X_{j} \) ~ Poisson i.e. \( X_{j} \) is a count | \( X_{j} \), i.e. the predictor itself |