SAS AutoModel takes a SAS file, constructs new derived variables, performs variable selection, and generates a logistic regression model. Thousands of variables and tens of thousands of observations can be handled, all automatically.
Key components, all written in SAS, include:
- "First look" display of graphical univariate information, including most frequent values for each variable with
- Automatic generation of cuts (x >= value) using standard information-theoretic decision tree criteria or proprietary
new Asymmetric Cut Factors to produce additional derived variables for modeling.
- Automated variable reduction to reduce the number of variables from as many as 10,000 to 100.
- Logistic regression modeling with standard and proprietary lift tables and measures of fit.
The Multi-Drug Modeler integrates multi-drug predictions with optimization capabilities for use by researchers, managed care organizations, physicians, and pharmaceutical companies.
The MDM is designed to:
Build single-drug and multi-drug predictive models for efficacy and side effects.
Compute optimized drug combinations, dosages, and schedules, taking into account individual patient limitations.
Incorporate genomic information from gene chips into predictive models and into optimization calculations.
The optimized combinations produced by the software is directly applicable to complex trials for multi-drug combinations. For inputs, MDM takes currently available data from a trial, plus contemplated experimental combinations that have not yet been attempted. The program then constructs a predictive model, based upon all of the currently known experimental data, and predicts results for all of the contemplated trials. This permits researchers to "tune" their explorations of the data during the course of their experiments, achieving cost reductions and finding better solutions.