Quantitative/ Analysis
Logistic regression
Model binary probabilities through log odds
Binary logistic regression models the log odds of an event as a function of predictors. Exponentiated coefficients are conditional odds ratios; fitted probabilities often communicate results more clearly. Multinomial and ordinal logistic models are related extensions with distinct outcome structures. Sparse events, separation and nonlinear predictors need explicit attention.
WHEN IT FITS
Use for a binary outcome such as default, retention or symptom status. Establish event definitions, sampling design and sufficient information across predictor patterns, especially when events are rare.
Strengths
- Produces probabilities constrained to the valid range
- Supports continuous and categorical predictors together
Limitations
- Odds ratios are often mistaken for risk ratios
- Separation and rare outcomes can destabilise estimation
Know the boundary
An adjusted odds ratio is not generally a risk ratio or a causal effect.
USED ACROSS
Banking & financeBusiness & MBAPsychologyComputer scienceEducation