Multilevel
Regression Analysis/Model are statistical models of parameters that
vary at multiple levels. The log odds of the binary outcome variables
are modeled as a linear combination of the predictor variables. This is done
when the data have fixed/random effects or are clustered. The Colley et
al 2016 paper used multilevel models with cross-level interactions to identify
variation in racial differences in low birth weight outcomes across
neighborhood risk levels when controlling for maternal demographics and
pregnancy behaviors.
Generalized estimating equations. These methods were utilized in the Headen article and estimate generalized linear model parameters with potentially unknown relationships between outcome measures. In the Headen article, log functions with generalized estimating equations were used to determine the risk ratios for gestational weight gain and race that reflected clustered pregnancies within various groups of women. Black and Hispanic women who were considered both normal and underweight experienced insufficient gestational weight gain with an increased risk compared to women that were White.
Mediation analysis was done in the Lorch article which demonstrated that SES factors among Hispanic mediated the disparities in 36 percent of fetal deaths. Mediation analysis is a model that aims to identify the underlying mechanism of a hypothesized causal chain in which one variable affects a second variable which can then affect a third variable (mediator) and can be statistically implemented by the use of sequential logistic regression models.
2. Describe a potential effect modifier, mediator, or contextual variable (for definition of contextual variable, see first page of option Merlo reading) for an association of interest to you and relevant to health disparities. For example, for investigating the association between SES and maternal mortality, I might be interested in the contextual variable of exposure to violence in the neighborhood. Describe how you would study whether this relationship exists.
A meditating variable may explain the relationship between an independent variable, such as maternal depression, and the dependent variable, such as childhood development/attachment, by another factor such as martial quality and or social support. In Taraban et al (Infant and Child 2017) they estimate the association between recurrent maternal depression and childhood development and attachment by examining the mediating role of social support and marital quality on childhood development. They surprisingly found that the association between maternal depressive symptoms and reduced parenting quality was strongest in the context of high marital quality and high social support, and largely nonsignificant in the context of low marital quality and low social support. The authors conclude that these results point to the importance of accounting for factors in the broader family and social context in predicting the association between depressive symptoms and maternal parenting.