Regression - Confidence Interval Calculation

michaelgloven

There is a new chart in RapidMiner Go showing actuals vs. predictions and a confidence interval. How is CI & margin of error calculated? Is it based on standard error of all the coefficients at 95% confidence level? Just looking for the underlying math. thanks!

itenger

Hi Michael,

RapidMiner Go draws the 95% confidence interval based on the absolute error variance returned by the Performance(Regression) operator in Studio Core. The operator is documented here: https://docs.rapidminer.com/latest/studio/operators/validation/performance/predictive/performance_regression.html 

If you want to know exactly how that variance is calculated we need to loop in a data scientist like @mschmitz

sgenzer

hi @michaelgloven I've pinged the RM Go team internally to get an answer for you.

Scott

michaelgloven

thanks Scott, looking forward to their reply.

itenger

Hi Michael,

RapidMiner Go draws the 95% confidence interval based on the absolute error variance returned by the Performance(Regression) operator in Studio Core. The operator is documented here: https://docs.rapidminer.com/latest/studio/operators/validation/performance/predictive/performance_regression.html 

If you want to know exactly how that variance is calculated we need to loop in a data scientist like @mschmitz