English to Portuguese Social Sciences Mathematics & Statistics religião, política, estatísticas

The analysis in this paragraph results from three multinomial logistic regression models regressing self-reported first round vote choice on reported clergy endorsements, controlling for religious affiliation.
In 2008, due to perfect multicollinearity, four individuals are excluded who said that both they and their pastors had supported the minor candidate [name]. The models from BEPS 2010 and 2014 are based on the post-election waves, and include a control for the lagged vote choice. The lagged dependent variables improve causal inference. Nonetheless, there are reasons to suspect that these observational results fail fully to capture clergy influence.