New method beats bias in censored data, improving accuracy and reliability.
Data censoring can mess up linear models, but the Tobit estimator can help if errors are normally distributed. However, if errors are not normal or have different variances, Tobit won't work well. Other estimators like CLAD and SCLS can handle this, but SCLS struggles with uneven errors. A new approach called partially adaptive estimation can deal with different variances and non-normal errors. In simulations, this new method works well with normal errors and beats Tobit and other methods with non-normal errors and varying variances. An example study supports these findings.
