Hoeffding's Inequality


Hoeffding's Inequality is a piece of math that is used to calculate how far a true, underlying, statistic of a population could deviate from some observed value collected from an IID sample under a given confidence interval.


The confidence interval

The expected value (E[x])

Variable range width


The x axis lower and upper limits


number of samples >= log(2/confidence interval)/(2*deviation from expected value^2)




The number of samples you have

The expected value (E[x])

Variable range width



The x axis lower and upper limits


confidence interval = 2*e^(-2*number of samples*deviation from expected value^2)