|
Visualizing Collinearity and Orthogonalization in Subject Space |
|
|
Visualizing Collinearity and Orthogonalization in Subject Space |
Date: 1998-1999Developers: Chong Ho Yu, Sandy Andrews, David Winograd, Natalie Schrodeder
Software: Director, Strata Instant Replay, Media 100, Photoshop
Media: Interactive animation
Description: Collinearity is commonly overlooked by many statistics students. Collinearity diagnostics such as variation inflation factor and tolerance index are too abstract for intermediate statistics learners to comprehend. To alleviate this problem, we developed a multimedia program to visually illustrate how a regression model collapses when the predictors are both mathematically and logically inter-related, as well as how the variance is artificially inflated by an overspecified model. The program also show how centering scores can overcome collinearity in a regression model with an interaction term.
You can download the beta program for evaluation. The file size is large. Don't do it at home unless you have a cable modem:
- Mac version (23 meg): This is a binhexed self-extracting archive. Your browser will automatically convert the file.
- PC version (17 meg): This is a zip file. You need Winzip to decompress the file.