Weight of Evidence (WoE) and Information Value (IV) are two concepts used in Logistic Regression to select variables and transform them. Here's how they are calculated:
WoE
Measures how much evidence supports or contradicts a hypothesis. It's calculated by taking the natural logarithm of the ratio of the proportion of responders to the proportion of non-responders. The value of WoE depends on the target variable's value, whether it's an event or a nonevent. The WoE will be zero if the odds of distribution goods to distribution bads are equal to one. If the distribution of bads is greater than the distribution of goods, the WoE will be negative. If the distribution of goods is greater than the distribution of bads, the WoE will be positive.
IV
A data exploration technique that determines which variable in a data set has predictive power. It's a numerical value that measures the overall predictive strength of an independent variable in capturing a binary dependent variable. IV is calculated by adding the absolute values of WoE across all groups. The value of IV can range from zero to infinity, with higher values indicating stronger predictive power.
