My Research

Research in Model Fitting Methods

Summer 2024

Kyle Smeedge, Zane Collis and I worked along side our mentor Dr. Jane Friedman to explore methods for model fitting piecewise linear models. We were pretty easily able to fit these piecewise linear models, but we didn't have a method for determining which fit was best. We couldn't simply take the fit with the least error, because models with more parameters tend to give less error. We came up with a clever way of comparing models side by side, and have the fit warp between models. We then used these comparisons to generate a decision vector whose direction and magnitude gave us a model and it's confidence.