Abstract by Angela Teuscher
Predicting House Prices using Gamma Regression
Gamma regression allows us to account for skewness in response variables without transfor- mations and is often more robust than other regression methods. One application of gamma regression is in predicting housing prices, which are strongly right-skewed. A model that predicts house prices well could be a useful tool for a website such as Zilllow. The final model used 11 predictors and the square root link, but had a root predictive mean squared error (RPMSE) of about $146,000 and simulated residual plots indicated a poor fit. Overall, the model is not good enough to be used as a prediction tool, but additional variables and a focus on middle class houses could result in a more appropriate model.