LEGIT (version 1.4.1)

example_2way_lme4: Simulated example of a 3 way interaction GxExZ model

Description

Simulated example of a 3 way interaction GxExZ model (where G, E and Z are latent variables). $$g_j \sim Binomial(n=1,p=.30)$$ $$j = 1, 2, 3, 4$$ $$e_k \sim Normal(\mu=0,\sigma=1.5)$$ $$k = 1, 2, 3$$ $$z_l \sim Normal(\mu=3,\sigma=1)$$ $$l = 1, 2, 3$$ $$g = .2g_1 + .15g_2 - .3g_3 + .1g_4 + .05g_1g_3 + .2g_2g_3$$ $$e = -.45e_1 + .35e_2 + .2e_3$$ $$z = .15z_1 + .60z_2 + .25z_3$$ $$\mu = -2 + 2g + 3e + z + 5ge - 1.5ez + 2gz + 2gez$$

\(y \sim Normal(\mu=\mu,\sigma=\code{sigma})\) if logit=FALSE

\(y \sim Binomial(n=1,p=logit(\mu))\) if logit=TRUE

Usage

example_2way_lme4(N, sigma = 1, logit = FALSE, seed = NULL)

Value

Returns a list containing, in the following order: data.frame with the observed outcome (with noise) and the true outcome (without noise), list containing the data.frame of the genetic variants (G), the data.frame of the \(e\) environments (E) and the data.frame of the \(z\) environments (Z), vector of the true genetic coefficients, vector of the true \(e\) environmental coefficients, vector of the true \(z\) environmental coefficients, vector of the true main model coefficients

Arguments

N

Sample size.

sigma

Standard deviation of the gaussian noise (if logit=FALSE).

logit

If TRUE, the outcome is transformed to binary with a logit link.

seed

RNG seed.

Examples

# Doing only one iteration so its faster
train = example_2way_lme4(250, 1, seed=777)
D = train$data
G = train$G
E = train$E

F = y ~ G*E
fit = LEGIT(D, G, E, F, lme4=FALSE, maxiter=1)
summary(fit)
F = y ~ 1
fit_test = GxE_interaction_test(D, G, E, F, criterion="AIC", lme4=FALSE, maxiter=1)
fit_test
#fit_test = GxE_interaction_test(D, G, E, F, criterion="cv", lme4=FALSE, maxiter=1, cv_iter=1)
#fit_test

F = y ~ G*E + (1|subject)
fit = LEGIT(D, G, E, F, lme4=TRUE, maxiter=1)
summary(fit)
F = y ~ (1|subject)
fit_test = GxE_interaction_test(D, G, E, F, criterion="AIC", lme4=TRUE, maxiter=1)
fit_test
#fit_test = GxE_interaction_test(D, G, E, F, criterion="cv", lme4=TRUE, maxiter=1, cv_iter=1)
#fit_test