Typically there is a twofold aim in the conduct of epidemiological analysis; (i) to estimate the average total causal effect and (ii) to disentangle the pathways which link the exposure and the outcome. When the causal effect is separated into the effect of the exposure through the mediator, it is referred to as indirect effect and the effect of the exposure that is not explained by the mediator is referred to as the direct effect. Traditionally, product and difference methods are used for estimating the direct and indirect effects. Methods using counterfactual framework have been proposed to accommodate interactions as well as extend the analysis for different types of variables (e.g. binary mediator, multinomial exposure). Under certain assumptions of confounding the direct and indirect effects estimated using counterfactual theory have causal interpretations. Traditional and modern methods allow estimation of path specific effects upon combining arbitrary models for mediator and outcome. Alternatively, methods have been proposed to estimate these effects using a single model, referred to as natural effect models. The aim of this study is to show applications of the natural effect models and demonstrate how the estimates from these models are doubly robust.