Detection and treatment of outliers in survey data

Abstract

In this thesis, we review the most common methods that are used for the detectionand treatment of outliers in survey data. The term outliers, in survey sampling,refers to sample units that have either a high value or are associated with a largesampling weight, thus having a high impact on the estimates. There is a varietyof methodologies that can be applied to survey data and which are presented inthe following chapters of this thesis. However, only few of those methods aresuitable for detecting and treating outliers when the survey design is complex. Inthis thesis, some of the methods are applied in an empirical study of outliers in thedata of the Survey on Income and Living Conditions of Greece, for the year 2013.