Analyses and results. Chapter IV Before turning to the regression analyses, this paragraph describes the explanatory variables, as well as the bivariate relationship between them and the dependent variable. Firstly, groups on average invest somewhat more in organizational capacity than in mem- ber involvement. More specifically, as shown in Table A2 in Appendix to Chapter IV, groups score 1.486 (S.D. 0.643) and 1.787 (S.D. 0.481) for member involvement and organizational capacity respectively. In addition, 37% of the interest groups in the analysis 88 are organizationally prepared to function as transmission belts. Secondly, regarding the bivariate analysis, the three explanatory variables are positively related to the degree of access interest groups gain to Commission officials. However, as shown in the in the correlation matrix (see Appendix to Chapter IV), the only variable that is significantly related to the degree of access is organizational capacity. Regarding the multivariate analyses, Table 4.2 presents the results of negative binomial regressions that have as dependent variable the level of access, that is, the number of meetings with public officials of the Commission. This is an appropriate method because the dependent variable is an over-dispersed count variable that ranges from 1 to 116, with a mean of 8.204, and a standard deviation of 15.431 (Long & Xxxxxx, 2014).
Analyses and results. Table 1 presents means, standard deviations, internal consistency estimates, and correlation coefficients for all study variables. We tested our theoretical model using path analyses in XXXX 23 (Xxxxxxxx, 2014). Our final sample size for the path analysis, conducted with full information maximum likelihood (instead of using methods like listwise and pairwise deletion, or mean im- putation), was 162. We specified covariances between the exogenous variables in our model. In addition, we allowed for a covarying association between intrapersonal and interpersonal need fulfillment because we consider these distinct yet related constructs. To assess fit to the data, we report the chi-square value (χ2), the Comparative Fit Index (CFI; Xxxxxxx, 1990), the Normed Fit Index (NFI; MacCallum, Roznowski, Mar, & Xxxxx, 1994), and the Root-Mean-Square Error of Approximation (RMSEA; XxxXxxxxx, Xxxxxx, & Xxx, 0000). The mediation hypotheses were tested with a package called ‘RMediation’ (Tofighi & XxxXxxxxx, 2011). The program produces estimates of indirect effects as well as confidence intervals around such effects on the basis of the distribution-of-the-product method. First, we tested our hypothesized model (Model 1) that integrates all the relationships proposed in the hypotheses. Although the chi-square test was significant (χ2(16) = 27.37, p = .038), other fit indices further showed that this model fit the data rather well (e.g., NFI = 0.95, CFI = 0.98, RMSEA = 0.07), which suggests general support for the hypotheses advanced herein. Nevertheless, we also tested alternative structural path models and Table 2 summarizes results comparing the different nested models through the Δχ2 statistic. Model 1 is the hypothesized model and contains only the hypothesized paths. Compared to the hypothesized model, Model 2 contains additional paths from interpersonal justice and perceived coworker support to intrapersonal need fulfillment and from procedural justice and job autonomy to interpersonal need fulfillment. Building on Model 2, Model 3 includes additional paths from intrapersonal and interpersonal need fulfillment directly to OCBs. Compared to the hypothesized model, Model 4 contains two additional paths from intrapersonal and interpersonal need fulfillment directly to OCBs (but not the paths from interpersonal justice and perceived coworker support to intrapersonal need fulfillment and from procedural justice and job autonomy to interpersonal need fulfillment as Models 2 and 3 contain). Com...
