Practical Predictive Analytics

Distribution of average errors by group

Distribution of errors is another that you can look at how well a model has fit the data. In this analysis, we look at the distribution of errors for all four combinations of the following variables:

• Outcome
• Correct prediction flag
• Average error as grouped into categories via the errbin variable:

result2 <- sql("select outcome,correct,errbin, count(*) as k, avg(error) as avgerr from preds_tbl group by 1,2,3 order by 1,2,3")