Closed test algorithm for choosing a fractional polynomial model with maximum permitted degree of 2 for a single continuous predictor. The first step is to determine whether a predictor should be included in a model. That is to compare models with and without FP2. If FP2 model is not better than null model, the predictor is dropped. Otherwise, we continue to compare FP2 with linear model. If FP2 is not better than linear one, we choose linear model. Otherwise, we continue to compare FP2 with FP1. If FP2 is not better than FP1, the FP1 model is chosen. Otherwise the FP2 model is chosen. FP, fractional polynomial.