A conceptual framework to deal with outliers in ecology

Abstract

Research on ecology commonly involves the need to face datasets that contain extreme or unusual observations. The presence of outliers during data analysis has been of concern for researchers generating a lot of discussion on different methods and strategies on how to deal with them and became a recurrent issue of interest in debate forums. Systematic elimination or data transformation could lead to ignore important ecological processes and draw wrong conclusions. The importance of coping with extreme observations during data analysis in ecology becomes clear in the context of relevant environmental aspects such as impact assessment, pest control, and biodiversity conservation. In those contexts, misinterpretation of results due to an incorrect processing of outliers may difficult decision making or even lead to failing to adopt the best management program. In this work, I summarized different approaches to deal with extreme observations such as outlier labeling, accommodation, and identification, using calculation and visualization methods, and provide a conceptual workflow as a general overview for data analysis.