“…The option to construct the estimator as a cellwise robust M regression as opposed to alternative paths, such as MCD regression (Rousseeuw, 1984), comes from the observation that robust M regression estimators have proven to yield a very good trade-off between efficiency and robustness in simulations and applications in fields as di-verse as quantitative structure-property relationships (QSPR) (Serneels et al, 2006), gravimetry (Hu et al, 2017), marketing (Guerard, 2016), chemometrics (Hoffmann et al, 2015), analytical chemistry with applications to e.g. analysis of archaeological glass (Serneels et al, 2005) and meteorite samples (Hoffmann et al, 2016), as well as estimation of shaping coefficients for futures trading in the electricity markets (Leoni et al, 2018). Note though, that S-regression has also proven a valid path in this context ( Öllerer et al, 2016).…”