Indexes of Validity and Reliability for Cross-Societal Measures

Abstract

Extracting measures from classical ethnographies has always necessitated proxy indicators in cross-societal research. Methods are given for calculating indexes of validity and reliability for both a low inference measure x and high inference proxy measure y of the same conceptual variable Z. Where two different ethnographic descriptions A and B exist for each society in a sample, low inference measures x _A and x _B are used to assess validity and reliability from the classical “true score” measurement model x = τ + ε _X , τ is the true score, ε _X is error, ξ[E _X ] = 0. Tests establish whether the model fits data. Reliability is defined to be (1 — Var[E _X ]/Var[X]). Validity is Var[T]/Var[X]. The validity and reliability of high inference measure y are increased by fitting a curve x = f(y) + ε _f(y) , ξ[E _f(y) ] = 0, to data. Standard cross-cultural sample examples are given. Solutions for cross-societal research may be considered for cross-cultural research.