freedom) decreased from 2.35 to 2.30, indicating an overall better goodness of fit of the reconfigured model compared to the original model. The squared multiple correlations value also increased slightly from .52 to .55, indicating a strengthening of the model’s measurement of the construct.
In summary, the goodness of fit measurements significantly improved after reconfiguration of the structural equation model to include only those variables which were statistically significant predictors of homophobia (age, interpersonal contact with gay men and lesbians as friends/family, belief in the “free choice” model of homosexuality, and support or non-support of a workplace nondiscrimination policy protective of gay men and lesbians). However, it must be stressed that models with many exogenous (independent) variables (such as the original model used in this study) are many times deemed unfit and elimination of insignificant variables reduces the overall number of variables in the SEM, thus, improving overall goodness of fit (Garson, 2005).
The research design employed in this study was purely quantitative. However, some qualitative trends in the analyses were observed due to free responses provided by some of the nurses within the sample (although the survey instrument has no questions requesting a free response from the participants). Of the 165 surveys included in this study, 16 had personalized comments hand-written by the participant on the survey