The Model Analyst's Toolkit: Scientific Model Development, Analysis, and Validation
Abstract
While there are various software tools that are used by the scientific community to help them develop and analyze their models (e.g., Excel, R, Simulink, Matlab), they are largely so general in purpose (e.g., Excel, R) or so focused on computational models in particular (e.g., Simulink, Matlab), that they are not ideal for rapid model exploration or for use by non-computational scientists. They also largely ignore the problem of validating the models, especially when the models are positing causal claims as most interesting scientific models do. To address this gap, Charles River Analytics undertook the "Validation Coverage Toolkit for HSCB Models" project with ONR. Under this effort, we successfully designed, implemented, informally evaluated, and deployed a tool called the Model Analyst's Toolkit (MAT), which focused on supporting social scientists to visualize and explore data, develop causal models, and validate those models against available data. During the current reporting period, we focused primarily on two areas. First, we released a new version of the MAT software, so much of the period was devoted to bug fixing, UI improvements, testing, and updating the user manual. Below, we summarize some of the key efforts in that area, and the new user manual is attached as an appendix. Second, we continued to focus on publishing about MAT and getting the word out to the research community. Section 5 describes our efforts in that regard.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 20, 2014
- Accession Number
- ADA616414
Entities
People
- Scott N. Reilly