The odds are linked to an

The odds are linked to an array of explanatory variables, as explained in (2): It is noteworthy that the pattern of parallel lines, shows the distributional p2x receptors that there is no individual difference between the ordinal J categories of the dependent variable, which refers to the value of β and Statistical significance. That is, the standard model assumes that the coefficients (β) keeps the same over existing category, this hypothesis is called Hypothesis of Parallel regressions for any ordered model. However, a problem with this proposal is that it invariably is violated, that is, commonly one or more of the β’s differ, making the parallel lines model restricted. Regarding the general ordered model, it makes flexible the constancy of the coefficients from the equations, and can be written as: In practice, the generalized model has the difference of the hypothesis that the estimated β’s are divergent for all J categories of the dependent variable. According Long and Freese (2006) and Williams (2006) the equations of the formulas of the parallel lines model and of the generalized ordered model are the same, except that in the parallel lines ordered logit, the β’s match with the j\'s, but not the α’s. The other option mentioned for the occurrence of violation of the hypothesis of parallel regressions is the ordered logit model of partial proportional odds. The model in question is a more parsimonious alternative between the two models already presented, being an intermediate model between the standard and the generalized models, as in this case, as Williams (2006) we have a partially restricted generalized ordered logit; when some β’s are the same for all values of j, but others are free to vary. Putting into an equation we have: To choose one of the three models to adopt, the Brant test (1990) was used as a criterion and indicated the ordered logit model of partial proportional odds as the most appropriate (see Table 6). Therefore, the ordered model used in this research will estimate the coefficients and the marginal effects of the frequency measure to services or religious activities in Brazil, using socioeconomic, demographic and cultural indicators.
Analysis and discussion of estimated results This study aims to identify the determinants of religious frequency of women in Brazil. For this, an ordered logit model of partial proportional probability is estimated and the results are reported in Table 5. It should be noted that the table reports the estimation of two models. In one of them DNA replicase is considered the religion in which the woman was educated (Catholic, Evangelical and none), while the other includes information about the current religion. The results analysis is derived from interpretation of the marginal effects. In a study of religion and fertility among adolescents in Brazil, Verona and Dias Júnior (2012) also find that the Protestant women aged 15–24 years old attend church more assiduously than the Catholic. The educational level has a positive correlation with religious frequency of Brazilian women, which is contrary to the study of Moreira-Almeida et al. (2010) for Brazil, and Halman and Draulans (2006) for European countries. However, it is observed that the effect is small, considering that the higher the educational level (years of schooling) of the Brazilian woman, it is more likely that she attend church once a week, 0.9 p.p Also note that the higher the educational level, it is less likely that woman never go to church (12.49 p.p). With regard to marital status of Brazilian women, one can say that the fact of being married or does not affect church attendance. Note that there is no statistical significance of the married variable with any level of frequency, in other words there is no consistent association between marital status and religiosity of Brazilian women. It is noteworthy that Moreira-Almeida et al. (2010) also did not observe any association. Another feature that did not show any effect on frequency was the fact of having a son or daughter that died. This result does not corroborate the description of Morelli et al. (2013) for whom spirituality and religiosity are perceived as coping strategies to obtain comfort, relief from suffering and acceptance.