WOLFRAM

Wolfram Archive

RiskQ 4.2 Helps Users Make Accurate Predictions

Published March 31, 2003

March 31, 2003–The new RiskQ 4.2 application package allows Mathematica users in finance, engineering, medicine, and science to use observational data–including time series–to develop data models, predict events, and create customized statistical tests and then to plan accordingly.

A recent article in Harvard Business Review demonstrates how a manager creating a budget by using only the average demand would systematically underestimate the cost of inventory shortfalls and surpluses. The article suggests that he should have used a more descriptive model for his data.

With RiskQ, users can turn their data into empirical distribution functions that can be directly manipulated by Mathematica. Because of Mathematica‘s sophisticated ability to handle symbolic operations, users can incorporate these distributions into their models as easily as they would plug in single values. With these distribution functions, users can do hypothesis testing; calculate statistical measures such as standard deviations, quantiles, and variances; and run simulations.

With Monte Carlo simulations, users can obtain frequency curves resulting from the interacting variations of distinct trials. RiskQ offers users the option of dramatically improving on the performance of their Monte Carlo simulations by using an efficient constrained Latin-Hypercube algorithm. This algorithm ensures that events dependent on the tails of distributions are adequately studied and leads to more accurate predictions of rare but cataclysmic events–for example, in estimating the risk of complete failure for a mission-critical system.

Technical features of RiskQ include a variety of parametric and nonparametric statistical functions that can be used to generate statistical analysis for means comparisons, distributional comparisons (Kolmogorov-Smirnov type tests), multivariate regression, correlation analysis, ANOVA/MANOVA, and homogeneity of one-way or multiway classified data.

Intimate knowledge of Mathematica is not necessary to use RiskQ although more powerful applications of RiskQ will certainly benefit from such knowledge. In addition to detailed descriptions of specific RiskQ functions, the manual includes a tutorial to introduce readers to useful Mathematica procedures such as creating symbolic functions, controlling numerical accuracy, and conducting simultaneous operations on all elements of a list or table.

“It’s time for a shift in mind-set,” states Harvard Business Review. “Rather than ‘Give me a number for my report,’ what [decision-makers] should be saying is ‘Give me a distribution for my simulation.'” With RiskQ 4.2 and Mathematica, users can build their simulations–and conduct many other types of tests–more easily than ever before.

More information about RiskQ is available.