Censored Regression

Most � vs a data are censored.  Censoring means that the true value of the signal is not known other than it is less than some noise threshold (left-censoring) or greater than 100% output, i.e. signal saturation, (right-censoring). Using ordinary least-squares regression with censored data will produce a grossly skewed result so special methods are needed.

(Censored regression is also mandatory for the proper statistical treatment of "runouts" - tests that are stopped before completion - for example in metal fatigue testing.)