Skip to contents

Example for XMInt

Source: vignettes/Example.Rmd
Example.Rmd

This package conducts model selection for multivariate mediation analysis involving exposure by mediator interaction while addressing the underlying hierarchical structure between the main effect and interaction.

Load package

Data generation

Example data (exposure X, outcome Y, and potential mediators M) was generated as follows.

For each subject \(i = 1, ..., N\):

  • \(X_i \stackrel{i.i.d}{\sim} N(0, 1)\)
  • \(M_{i,v} = a_v X_i + e_{1_{i,v}}\), where \(e_{1_{i,v}} \stackrel{i.i.d}{\sim} N(0, 1), v = 1, ..., V\)
  • \(Y_i = X_i + \sum_v b_{1_{v}} M_{i,v} + \sum_v b_{2_{v}} X_i \times M_{i,v} + e_{2_{i}}\), where \(e_{2_{i}} \stackrel{i.i.d}{\sim} N(0, 1)\)

The first three M variables (\(M_1,M_2,M_3\)) are set to be the true mediators (i.e., having non-zero \(a\) and \(b_1\) coefficients), \(X \times M_1\) is set to be the true exposure-by-mediator interaction term (i.e., having non-zero \(b_2\) coefficients), and all other coefficients are set to be 0. The effect size (ES) represents the value of \(a, b_1, b_2\) of the truth, which are \(a_1, a_2, a_3, b_{1_{1}}, b_{1_{2}}, b_{1_{3}}, b_{2_{1}}\) in our case.

## data generation
data = dat_gen(N = 200, V = 50, es = 1, seed = 1)
X = data$X; Y = data$Y; M = data$M

Model selection 

## select model
result = XMInt_select(X,Y,M)

Results 

## selected mediator(s)
result$selected_mediator
#> [1] "M1" "M2" "M3"

## selected interaction(s)
result$selected_interaction 
#> [1] "M1"