Some evidence shows it does not, at least when examining the relationship between variables in a multivariate model, though response-rate bias is found for univariate distributions (Rindfuss et al. 2015).
In cases where the genetic variant contributes to only one trait, a loss of power will occur with the multivariate model, though the magnitude of this loss is moderate.
Obesity remained significant in the multivariate model, though with a decreased OR, probably influenced by Low education, which is a risk factor for obesity.
Women in spontaneous or induced labor had also an excess risk of 26%and38%8% respectively, while women with elective CS had an excess risk of 50% after introducing socio-economic characteristics in the multivariate model, though not statistically significant.
The mental health variables studied, with the exception of self-esteem, were not significantly related to self-rated health in the multivariate model, though several were significant in two-way tables.
Both education and occupation showed significant impacts on household income in 2008 (both p values ≪ 0.001 in a generalized linear model), and household income was selected for the multivariate model even though its univariate p value was ≫ 0.25.
The risk of resistance to any antiretroviral class was not influenced by the use of TA-including or sparing regimens or by the use of FTC rather than 3TC in the multivariate model, even though a lower prevalence of any NNRTI resistance was found in subjects assuming FTC compared to 3TC (52.6% vs. 61.2%, p = 0.012).
Though problems due to some autocorrelation were expected, the linear multivariate models performed quite well.
In multivariate models, among non-pregnant women, VAW was significantly associated with all gynecologic outcomes, though not with self-rated health (Table 4).
All predictors and covariates were entered simultaneously in the final multivariate models.
