Processing Data for Outliers

摘要: Every experimenter has at some time or other faced the problem of whether certain his observations properly belong in presentation measurements obtained. He must decide these are valid. If they not valid will wish to discard them least treat data a manner which minimize their effect on conclusions. Frequently interest this topic arises only final stages processing. It is author's view that consideration sort more made recording stage perhaps preliminary I This be discussed terms following general models. We assume independently drawn from particular distribution alternatively, we an observation occasionally obtained population and there nothing experimental situation indicate happened except what may inferred observational reading itself.2 if no extraneous occur, (or transformation them, such as logs) follow normal distribution. shall also occasional either with shifted mean same larger variance. These assumptions completely realistic but procedures developed for alternatives should helpful. one taking where models apply remain two distinct problems. First, attempt pick out different populations. One interested selection something gone wrong procedure resulting (in case he include result) gives indication unusual occurrence investigator explore further.