Validity and Reliability (chapter 5)
The interpretations of the results presented assume that the measures incorporated in the analyses are both valid and reliable. The models tested include four theoretically distinct dimensions (variables): PREMIUMS, GOALS, INTERACTIONS, and THREATS. Concern for the validity and reliability of these concepts is particularly important, given the motivational character of the variables and the fact that these theoretical concepts have not yet been analyzed empirically in this form.
Validity refers to the extent to which the measures of the variables correspond to the concepts they are intended to reflect. Generally, there are four basic approaches to validation. First, pragmatic validation, assessing how well the measure works in allowing one to predict behaviors and events, requires alternative indicators of the concepts that are themselves valid indicators. Here no such measures are available; unfortunately, there are few instances in social science research where valid alternative indicators exist. The second type of validation, construct validation, is more applicable to this analysis. There are two types of construct validation: internal or convergent and external. 1 Convergent validation involves devising several measures of the same variable and assessing the degree of correlation among the similar measures. Multiple indicators are employed for each concept. For the PREMIUMS, GOALS, and THREATS variables, three indicators were used. For the INTERACTIONS variable, only two indicators were available from the survey data. The correlation matrices for each concept are presented in table C.1. The average correlations are: PREMIUMS indicators (.415), GOALS indicators (.452), THREATS indicators (.396), and INTERACTIONS indicators (bivariate) (.494). All correlations are significant at.0001 level. These correlations suggest that the indicators used in the construction of each of the conceptual variables are