7. Hypothesis testing
While data for the potentially unethical behavior of buyers and suppliers was collected from 88 matched dyads, there are less than 88 matched data sets for all of the constructs included in H1a–H4, due to item non-response. Structural equation modeling is an appropriate analysis method for data in which a series of regression analyses are being performed, where the dependent variable for one regression analysis is also the independent variable for another et al., 1992 . However, as noted by Anderson and Gerbing 1988 , while, ‘‘the bias in parameter estimates is of no practical significance for sample sizes as low as 50,’’ improper solutions, including negative eigenvalues, can occur for samples of this size