| Modeling Statistical Phenomenon Using Interactive Simulation
Thom R. Williams
Department of Statistics
Colorado State University
Monday, October 17, 2005
Data Generator is a simulation tool that provides Instructors with an interactive platform to build a model that specifies desired statistical phenomenon, generate datasets based on the model for case studies and illustrations, and interact with students for practical and applied hands-on analyses to reinforce learning. The tool's interaction with the student has proven to be an invaluable resource, especially when the model requires the student to perform appropriate diagnostics and remedial action. The student can employ these real-time during in-class exercises or homework assignments. The tool interacts with the student using passive and active studies. For passive studies, the student inputs a desired sample size, and then the tool will generate data based on the Instructor's specified model. However, time and cost constraints are included with this simulation: each sample takes time and has an associated cost. This encourages the student to establish a sampling rationale. For active studies, the students can create a designed experiment and then use the tool to obtain the simulated results. Again, time and cost constraints are present, which should encourage a parsimonious design strategy.
Modeling capabilities for this tool are limited to linear models; however, the Instructor has the option to introduce the following:
- Specified level of ‘noise' for the error term
- Higher order terms
- 2-way and 3-way interactions
- Optimization studies with ‘competing' responses
This is the first generation for this tool. Future generations will include the introduction of time series, nonlinear models, discrete responses, and non-normal error terms.
[Mr. Williams' M.S. oral examination will immediately follow the seminar presentation.]