Webinar: Problems with R-squared in Model Building and Decision Making
In generalized terms, there are two types of statistical models we attempt to create. One is some form of a regression model, where we take historical data and attempt to model the independent variables (inputs) against a dependent variable (output) - however they might have behaved. The other is a model created from an experimental design where we carefully control the variation of the inputs to model the output. Both of these analyses utilize Analysis of Variance (ANOVA) and one of the statistics generated is R-Squared (as well as R-squared adjusted). The interpretation of what R-squared actually means in these two types of models is radically different. Furthermore, we will explore in the regression setting how businesses can make mistakes in decision making due to the fascination of “higher R-squared is better” in model building.