outlier effects

views updated

outlier effects The pattern created in a sample by the inclusion of a few atypical cases or groups. For example, if one finds that data from twenty different schools show there has been an overall decrease in the ratio of staff to students, it will be necessary to check whether this decrease is general or specific merely to one or two institutions. If the staff:student ratio has decreased dramatically in one school but not others, this idiosyncratic case could create what appears to be a general pattern, when combined with other schools in which there has been little or no change. These effects can also be observed in regression analysis, because outliers can distort regression lines and coefficients, so the prudent researcher usually checks for their presence by also displaying data fully in a contingency table or scattergram.