Graphical representation of how the discrimination slope for a time t is calculated using survival data. The discrimination slope (DS) is probably more well known as one half of the integrated discrimination improvement (IDI). Specifically, IDI is the improvement in DS between two survival models (typically model A uses a baseline set of predictors and model B adds a predictor of interest).

There is an interactive element: click on the desired time t on the scatterplot, which will trigger the animation for calculating DS at t. Click again to reset.

Note that this shows an inefficient estimation method, as censoring is ignored. It's useful mainly as an intuitive visual teaching tool. Uno et al. (2012) developed an estimator that incorporates censoring and assumes the Cox model. They also have an accompanying R package.

Some things yet to do:

• Still need to convert from native predictor scale to predicted probability scale
• Need to insert explanatory text and tooltips

Please feel free to suggest comments and/or submit pull requests! Built using D3.js.