*2.4. Statistical Analyses*

Analyses were carried out using IBM SPSS statistics version 21 (IBM Corp., Armonk, New York, NY, USA). The PROCESS macro version 3.1 [53] was used to test the study's hypothesized model, which included both a moderation (i.e., the individual's gender) and a mediation (i.e., sexual orientation concealment and internalized heterosexism) component.

A total effect moderation model (model template 59 in the PROCESS macro) was first specified. In this model, the effect of prejudice events (independent variable) on parenting desire (binary dependent variable; direct effect), and on sexual orientation concealment and internalized heterosexism (mediators), as well as the effect of both sexual orientation concealment and internalized heterosexism on parenting desire, were supposed to be conditional (or moderated) by the individual's gender (woman vs. man); however, estimation of the coefficients in the statistical model revealed no significant interaction between prejudice events and gender on both sexual orientation concealment and internalized heterosexism (both *ps* > 0.05). Since nonsignificant interactions also influenced the estimate of the hypothesized indirect effects (which were necessarily conditional with these interactions in the model), along with all inferential tests thereof, they were trimmed in the final moderated mediation model (Figure 1), and were thus expressed by only three interaction terms: (1) gender by prejudice event interaction on parenting desire; (2) gender by sexual orientation concealment interaction on parenting desire; and (3) gender by internalized heterosexism interaction on parenting desire (model template 15 in PROCESS macro).

Significant conditional direct and indirect effects were probed by using the pick-a-point approach. As a final step, in order to quantify the indirect effects as a function of the moderator, and to provide inferential tests for those conditional indirect effects, indices of moderated mediation were estimated for each of the hypothesized moderated mediation paths, as recommended by Hayes [53]. Bias-corrected bootstrap confidence intervals, based on 5000 resamples, were used as indicators of effect size. Confidence intervals that did not contain zero indicated a significant indirect effect via the specific mediator.

**Figure 1.** The hypothesized moderated mediation model. For reasons of simplification, control variables were not reported in the figure.
