Software

Clemens van Brunschot wrote a SAS macro for performing a sensitivity analysis based on a predictive model. For each predictor variable the score is calculated at percentiles 1, 5, 10, 25, 50, 75, 90, 95, 99 of the predictor. With a specified percentile as a reference point the change in score is analyzed. All other predictors are kept constant at a specified percentile level. All predictors must be numeric. They are treated as ordinal measurement level or higher. Dummy variables (dichotomies), a frequency variable, and a selection of cases are allowed. The predictive model can presumably be from any type, as long as the score is numeric. The use of the macro requires that you write a scoring macro (an example is provided).

URL: members.brabant.chello.nl/~c.vanbrunschot/macros.html E-mail: vanBrunschot@bigfoot.com