Binary Logistic Regression
Analyse the effect of several factors on a binary outcome (e.g. occurred/not, effective/ineffective). Fitted by maximum likelihood (IRLS), it outputs each factor's OR (odds ratio) with 95% CI, significance, and model fit indices. A core method for clinical prediction and risk-factor analysis.
① Paste data (with header)
First row is column names, then one case per row. The outcome variable must be 0/1 (0=did not occur, 1=occurred). You can paste a whole block from Excel.
How to use & methodology
What is logistic regression used for?
It analyses the effect of one or more factors (predictors) on a binary outcome (e.g. diseased/not, effective/ineffective, relapse/no relapse) and is the most common method for clinical risk-factor analysis and prediction models. Results are expressed as OR (odds ratio), giving each factor's direction and strength.
How do I interpret the OR (odds ratio)?
OR>1 means the factor is a risk factor (increases the chance of the outcome), OR<1 means a protective factor (decreases it), OR=1 means no association. A continuous variable's OR is the change in the odds of the outcome per 1-unit increase. To judge significance, see whether the 95% CI includes 1 — it is significant only if it excludes 1.
How do I prepare the data?
Arrange a table: first row is column names, one case per row. The outcome variable must be coded 0 (did not occur) and 1 (occurred). Predictors can be continuous (e.g. age, lab values) or already-coded categorical values. Select the whole block in Excel and paste.
What is the 'complete separation' warning?
When a predictor almost perfectly separates the outcome (e.g. all positive cases have a higher value than all negatives), the maximum-likelihood estimate diverges and the OR becomes extremely large and unreliable. This is common with small samples or an overly strong variable and is not a tool error. Increase the sample, merge categories, or reduce predictors.