condition: Computes mgm object conditional on a set of variables

In mgm: Estimating Time-Varying k-Order Mixed Graphical Models

Computes mgm object conditional on a set of variables

Description

The function takes an mgm object and a set of variables fixed to given values as input and returns the conditional mgm object.

Usage

condition(object, values)

Arguments

object	An mgm object, the output of the mgm() function.
values	A list, where the entry name indicates the column number of the variable that should be fixed, and the entry value indicates the value to which the corresponding variable should be fixed. See below for an example.

Details

The new conditional object still contains the variables that were fixed, however, they are not related to any of the random variables anymore. We kept the variables in the object to avoid confusion with variable labels and plotting. Also note that mgm() by default scales all Gaussian variables to mean=0, sd=1. Thus, fixed values should be selected based on the scaled version of variables.

Value

The function returns an mgm object that is conditional on the provided values. The new mgm object can again be used as input in predict(), print(), showInteraction(), etc.. Note that mgm() by default standardizes variables to mean=0, SD=1. Therefore, also the values one conditions on should be chosen on the scaled version of the variable to avoid extrapolating out of the dataset.

Author(s)

Jonas Haslbeck <[e-mail]>

References

Haslbeck, J., & Waldorp, L. J. (2019). mgm: Estimating time-varying mixed graphical models in high-dimensional data. arXiv preprint arXiv:1510.06871.

See Also

mgm

Examples

## Not run: 

# --- Create Mixture of two Gaussians ---

set.seed(1)
n <- 500
library(MASS)

# Component A
Sigma_a <- diag(2)
Sigma_a[1, 2] <- Sigma_a[2, 1] <- .5
Xa <- mvrnorm(n = n, mu = rep(0, 2), Sigma = Sigma_a)

# Component B
Sigma_b <- diag(2)
Sigma_b[1, 2] <- Sigma_b[2, 1] <- 0
Xb <- mvrnorm(n = n, mu = rep(0, 2), Sigma = Sigma_b)

data <- cbind(rbind(Xa, Xb), c(rep(0, n), rep(1, n)))
colnames(data) <- c("x1", "x2", "x3")


# --- Fit MGM ---

# with mgm
mgm_obj <- mgm(data = data, 
               type = c("g","g","c"), 
               level = c(1, 1, 2), 
               moderator = c(3), 
               lambdaSel = "EBIC")

# --- Condition on / fix values of variable 3 ---

# Fix x3=0
mgm_obj_x3.0 <- condition(object = mgm_obj, 
                          values = list("3"=0))
mgm_obj_x3.0$pairwise$wadj

# Fix x3=1
mgm_obj_x3.1 <- condition(object = mgm_obj, 
                          values = list("3"=1))
mgm_obj_x3.1$pairwise$wadj


## End(Not run)

mgm documentation built on April 4, 2025, 1:03 a.m.