Conceptual Model 1997 Community Policing Measurement Model Component Fit Th e data sources oﬀ ered multiple measures of COP training and citizen interaction, permitting two ﬁ rst-order latent constructs. Th ese ﬁ t the data well. Table 8.1 shows the training ( = .001) and citizen interaction ( = .05) constructs were both statistically signiﬁ cant based on their construct variances and z-statistics. With recruit COP training and group meetings serving as a scale to their respective constructs,1 all the measures were statistically signiﬁ cant at the p<0.001 level and, as expected, positive. Overall, the relationships between
the measures and their corresponding constructs were strong. Th e weakest was that of data accessibility for citizen interaction, but even here the standardized factor loading, which indicates the strength and direction of the construct-measure relationship on a scale of –1 to 1, was 0.4235, signiﬁ cant at p<0.001, and the construct explained about 18 percent of the variance, as indicated by the R-square value.