Statistical Model Sample Clauses

Statistical Model. The thesis uses logistic regression to analyze the influence of women’s empowerment and community group participation on stunting. Logistic regression is a statistical technique used to model the probability of discrete (binary or multinomial) outcomes. Logistic regression analysis provides more efficient and powerful insights into what attributes are more or less likely to predict an event outcome in a population of interest by estimating the probability of its occurrence. Binary logistic regression, utilized here, is a form of regression that is used when the dependent variable is dichotomous and the independent variables are continuous, categorical variables, or both. Modeling strategy This thesis presents the results of a secondary data analysis and thus the variable selection for model building was limited to variables collected during the endline evaluation of the Window of Opportunity program. Variables in the model were selected based on two main factors: a) key domains of women’s empowerment and social and biological characteristics identified in the literature as strong predictors of child nutritional status and b) availability of potential dependent variables in the data. Knowing the nature of the outcome of interest (dichotomous) the logistic regression appeared to be the logical option for analyzing the relationships between the variables of interest. In the model building process, we first checked the distribution of all variables (histograms for normality and scatter plots for outliers). Second, direct modeling was used to include available variables identified as potential predictors of the outcome. The final model retained variables that, when included in the model, increase the overall R-Square of the full model. Based on this criteria we obtained an R-Square of 0.84 (the highest R-Square possible based on available variables in the data) when we included maternal age, maternal education, group participation, child sex and women empowerment. There are several known and unknown potential factors that may confound the relationship between women’s decision-making and child nutritional status. Known factors include but are not limited to birth order, child age, breastfeeding, household hygiene, seasonality, household size, women economic control, wealth quintiles, women education and husband’s occupation. While we were able to control for some of these confounding factors (child sex, child age, maternal age and maternal education) using strat...