For illustration, we analyzed the t

For illustration, we analyzed the total effect moderationmodel, the most general of the eight models shown in Figure1. This model is represented by Equations 5 and 20, whichwere estimated using SPSS (Version 14.0; 2005, SPSSInc.). The regression module was used to estimate coeffi-cients for the full sample, and the constrained nonlinearregression (CNLR) module was used to estimate coefficients from 1,000 bootstrap samples. Unlike the regression module, the CNLR module contains an algorithm that draws bootstrap samples, estimates regression coefficients for each sample, and writes the coefficients to an output file. We used the default loss function of the CNLR module, which minimizes the sum of squared residuals, thereby producing OLS coefficient estimates.