Validity, Invalidity, and Reliability

Abstract

Sociologists almost always rely on fallible measuring instruments in attempting to estimate parameters. Recent work by Heise (1969) has shown how parametric estimates can be made with fallible data when one has three waves of panel data. However, when one has only crosssectional data, he must obtain multiple measurements on the same underlying true variable and use the correlations among the fallible measurements to make estimates of the parameters. Two different approaches