VALIDATING AND COMPARING SIMULATION MODELS USING RESAMPLING

—————————————————————————————————————

The three Excel workbooks that can be downloaded below demonstrate the bootstrap methods discussed in the article

“Validating and comparing simulation models using resampling” by R.C.H. Cheng

The workbooks address two problems using bootstrap resampling methods.

The first problem concerns the comparison of a data sample with a parametric model fitted to the data.

NormalFit&GoFTestData – This fits the normal distribution to a data sample and carries out a goodness-of-fit test of the model using the Anderson-Darling test statistic with critical values calculated by bootstrapping.

GammaFit&GoFTestData – This fits the two-parameter gamma distribution to a data sample and carries out a goodness-of-fit test of the model using the Anderson-Darling test statistic with critical values calculated by bootstrapping.

The second problem concerns the non-parametric comparison of two data samples.

EDFITAnalysisForJOS – This carries out a comparison of the EDFs of two (or more) samples using the EDFIT method. The workbook calls routines in the DLL macro: adfitDLL.dll. This needs to be downloaded and placed in the same folder as the workbook.

Details of the EDFIT method are given in

Cheng R C H and Jones O D (2004). Analysis of distributions in factorial experiments. Statistica Sinica, 14, 4, 1085-1103.

Lecture Notes on resampling methods including links to Excel Workbooks containing other examples involving non-linear regression, logistic regression, ANOVA can be found at

http://www.maths.soton.ac.uk/staff/Cheng/Teaching/GTPBootstrap/

R.C.H. Cheng

10 August 2006