The bias and skewness of \(L_ 1\)-estimates in regression (Q1095536): Difference between revisions

We consider the linear and nonlinear regression with shift parameter \(\alpha\) and other parameters \(\beta\). The unknown distribution of the residuals is not assumed to be normal or symmetric. Let (\({\hat \alpha}\),\({\hat \beta}\)) be the \(L_ 1\)-estimate. Simple and easily computed approximations are given for the bias and skewness of \({\hat \beta}\). For the linear model these are proportional to the 3rd moments of the `independent' or `x' variables. For a nonlinear model its linear component plays the role of these `x' variables, and a second term must be added proportional to the covariance of its linear and quadratic components. An extension to random `x' variables is given.