Description
Modelling the probability of a binary UX outcome, such as task success, conversion or churn, from several predictors, with results expressed as odds ratios.
Logistic regression models the log-odds of a binary outcome as a linear function of predictors, keeping predicted probabilities between 0 and 1. Coefficients exponentiate to odds ratios: the multiplicative change in the odds of success per unit increase in a predictor. It is the right tool for task completion, conversion, churn and error occurrence, where linear regression would give impossible predictions and invalid inference. Rigour comes from enough events per variable (traditionally at least 10), checking linearity of continuous predictors in the logit, multicollinearity and influential cases, assessing discrimination (AUC) and calibration (calibration plots or Hosmer-Lemeshow), and translating odds ratios into predicted probabilities or marginal effects that stakeholders understand.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Active
- Stage
- Evaluation, Post-launch
- Data
- Quantitative
- Effort
- Days
Steps
- Define the binary outcome precisely and ensure at least about 10 events per predictor in the smaller outcome class.
- Fit the model (SPSS LOGISTIC REGRESSION or statsmodels Logit) with predictors chosen from theory.
- Check linearity of continuous predictors in the logit (Box-Tidwell or spline terms), VIFs and influential cases.
- Report odds ratios with 95% CIs and compute average marginal effects or predicted probabilities for key scenarios.
- Assess discrimination (AUC; around 0.7 acceptable, 0.8 good) and calibration, preferably with cross-validation.
- For ordinal outcomes switch to PLUM (ordinal logistic) and for clustered users use mixed-effects logistic models.
Sources
- SPSS REGRESSION
- PLUM
- LOGISTIC REGRESSION procedures; Hayes PROCESS macro v4
- Field, A. (2024) Discovering Statistics Using IBM SPSS Statistics, 6th ed.
- Hosmer, D. W., Lemeshow, S. & Sturdivant, R. X. (2013) Applied Logistic Regression, 3rd ed.
- Sauro, J. & Lewis, J. R. (2016) Quantifying the User Experience, 2nd ed.
- The Inquiry Series, C8 — SPSS for Regression, Moderation & Mediation
- Factor and Discriminant Analysis (Sauro & Lewis, MeasuringU training)
- Practical Statistics for Data Scientists (Bruce & Bruce)
- Benchmarking the User Experience (Sauro)
Also known as
- Binary logistic regression
- Logit model
Categories
- Analytics & Behaviour
- Statistics & Modelling