Currently, evaluation of models performance on traning data does not include any actual statistical significance testing. While R^2 is useful in discussing how much variance in the data the model explains, it would be nice to have some significance testing which would be more definitive.
An ANOVA test can be done on our models (at least on multiple linear regression models) that give a nice picture of the performance of the model (including the significance of different coefficients in a linear model) and whether the models's predictive power is significant under some confidence bound.
Done when:
Modeling framework(s) produce ANOVA reports for (at least) linear regressions and (if possible) any non-linear regression models that are used (don't really have any)
Currently, evaluation of models performance on traning data does not include any actual statistical significance testing. While R^2 is useful in discussing how much variance in the data the model explains, it would be nice to have some significance testing which would be more definitive.
An ANOVA test can be done on our models (at least on multiple linear regression models) that give a nice picture of the performance of the model (including the significance of different coefficients in a linear model) and whether the models's predictive power is significant under some confidence bound.
Done when: