Clark, RobertKokic, Philip N.Smith, Paul A.2021-04-270306-7734http://hdl.handle.net/1885/231034Two alternative robust estimation methods often employed by National Statistical Institutes inbusiness surveys are two-sided M-estimation and one-sided Winsorisation, which can be regardedas an approximate implementation of one-sided M-estimation. We review these methods andevaluate their performance in a simulation of a repeated rotating business survey based on datafrom the Retail Sales Inquiry conducted by the UK Office for National Statistics. One-sidedand two-sided M-estimation are found to have very similar performance, with a slight edge forthe former for positive variables. Both methods considerably improve both level and movementestimators. Approaches for setting tuning parameters are evaluated for both methods, and this isa more important issue than the difference between the two approaches. M-estimation works bestwhen tuning parameters are estimated using historical data but is serviceable even when only livedata is available. Confidence interval coverage is much improved by the use of a bootstrap percentileconfidence interval.application/pdfen-AU© 2016 The Authors. International Statistical Review © 2016 International Statistical Institute. Published by John Wiley & Sons LtdBootstrapmean squared errorM-estimationmovement estimationinfluential valuesoutliersrobustnesssample surveyWinsorisationWinsorization201710.1111/insr.121772023-03-19