Efficient Designs for Judgment Analysis (Paperback)

In formulating a public policy, judgment analysis enables policy makers to assess and accommodate the relative importance of differing viewpoints and concerns of competing actors as an analytic decision aid. Among the important constraints that prevent judgment analysis from being widely applied to the policy formulation process is a methodological limitation inherent in judgment analysis: It requires too many scenarios to be judged in a single judgment task. Addressing this issue, this dissertation implemented two efficient design concepts - efficient plausible design and augmented representative design - suggested by McClelland (1999) as alternative design frameworks for judgment analysis that balance two conflicting principles: the principles of representative design and statistical efficiency. It also sought to derive a generalizable rule about the minimum number of cases needed to arrive at reliable conclusions about a judgment policy given the judgment task. Additionally, it tested the applicability of the bootstrap analysis as an alternative method to estimate the stability of the coefficients modeled for a judgment policy given the limited number of observations.