Model a yes/no outcome and report it the way clinical papers do.
Your browser does not support embedded video. Download the video (MP4) .
:::
The R code
Everything in the video, ready to copy:
# Logistic regression - odds ratios & ROC library (gtsummary)library (pROC)# Fit a logistic model for tumour response <- na.omit (trial[, c ("response" , "age" , "grade" , "stage" )])<- glm (response ~ age + grade + stage, data = d, family = binomial)# Adjusted odds ratios with 95% CI tbl_regression (fit, exponentiate = TRUE ) |> bold_p ()# ROC curve and AUC - how well the model discriminates <- roc (d$ response, fitted (fit), quiet = TRUE )auc (roc_obj)plot (roc_obj, col = "#2f6fed" , lwd = 3 , legacy.axes = TRUE )▶ Practice in the playground
When to use it. Use it for a binary outcome; exponentiate the coefficients to get odds ratios, and judge the model with the ROC curve and AUC.
Go deeper: the Clinical R Toolkit · read logistic regression, OR & ROC .
Next: Kaplan-Meier Survival Curves →