Here leads back to the home page.
Like any scientific endeavor mistakes occur that (unfortunately) become published. On this page you can find discussion of these mistakes and their corrections, or the relevant links to the corrigenda.
J. Chown and U.U. Müller (2018). Detecting heteroskedasticity in nonparametric regression using weighted empirical processes. Journal of the Royal Statistical Society, Series B, 80, 951-974. Preprint on arXiv: 1610.09139.
Please find the appropriate corrections here.
J. Chown (2016). Efficient estimation of the error distribution function in heteroskedastic nonparametric regression with missing data. Statistics and Probability Letters 117, 31-39. DOI:10.1016/j.spl.2016.04.009.
In this article I discuss estimation of the error distribution function in nonparametric regression with heteroscedastic errors. A referee kindly asked if the expansion of the residual-based empirical distribution function I provided could hold under more mild conditions on the error distribution function F than those of Neumeyer and Van Keilegom (2010), Estimating the error distribution in nonparametric multiple regression with applications to model testing, Journal of Multivariate Analysis 101, 1067-1078. The idea of lightening the conditions for the expansion to hold revolved around proving that the error density function f satisfied a crucial Lipschitz property when the Fisher information integral for location and scale is finite. My argument was malformed. Rather, Assumption (C5) of Neumeyer and Van Keilegom (2010) yields a correct expansion argument executed in the usual fashion (see the proof of their Theorem 2.1), and the conclusion of Theorem 1 of Chown (2016) remains unchanged, i.e. replace the Assumption 2 in that paper with Assumption (C5) from the previous authors. I apologize to the authors Natalie Neumeyer and Ingrid Van Keilegom for this error since I stated that the result incorrectly generalized their work.