Simulating mixed-effects longitudinal logistic data in R

Many thanks all!  I basically want to generate 100 bernoulli outcomes
whose
probability of being 1 is longitudinally modeled based on 4 time
points
(0:3), membership to 2 groups (Control vs. treatment), and gender
(male vs.
female).

Thanks a lot!

On Sun, Jul 12, 2020 at 5:02 PM Ben Bolker <bbolker at gmail.com> wrote:

   Or:


(1)  see ?lme4::simulate.merMod(), i.e. set up your covariates,
then
simulate(formula, newdata=data_frame, newparams=...,
family=binomial,
weights=1)

or

(2) after setting everything up as in your example, use

response <- rbinom(length(y), prob=plogis(y), size=1)

On 7/12/20 5:58 PM, D. Rizopoulos wrote:

Have a look in this example:

Best,
Dimitris

??
Dimitris Rizopoulos
Professor of Biostatistics
Erasmus University Medical Center
The Netherlands

behalf of Simon Harmel <sim.harmel at gmail.com>

Sent: Sunday, July 12, 2020 10:44:17 PM
To: r-sig-mixed-models <r-sig-mixed-models at r-project.org>
Subject: [R-sig-ME] Simulating mixed-effects longitudinal
logistic data

in R

Good afternoon,

I know to simulate a linear mixed-effects longitudinal data in R,
I can

use

the procedure below.

But I wonder how to generate a similar design, but for a logistic
model

in

R?

I appreciate your suggestions,
Simon

#------------- linear mixed-effects longitudinal data in R-------

library(MASS)

growth.sim <- function(J, n.time, gammas, G, sigma = 1, seed =
NULL) {

    X <- cbind(1, seq_len(n.time) - 1) # time indicators for each

individual

    X <- X[rep(seq_len(n.time), J), ]  # Repeat each row n.time
times

   st.id <- seq_len(J)                 # individual id
   st.id <- rep(st.id, each = n.time)  # repeat each ID n.time
times


set.seed(seed) ## what should go below, possibly correlated Beta?
uj <- MASS::mvrnorm(J, mu = rep(0, 2), Sigma = G) # Generate u0
and u1
random effects for intercepts and slopes

## I think no error term is needed for logistic version?
eij <- rnorm(J * n.time, sd = sigma)              # Generate
error
term for observations


betaj <- matrix(gammas, nrow = J, ncol = 2, byrow = TRUE) + uj #
Compute beta_j's

y <- rowSums(X * betaj[st.id, ]) + eij          # Compute outcome
based on Yij = sum(Xij*Bj) + eij:

dat <- data.frame(st.id, time = X[ , 2], y) # Output a data frame
return(dat)                                 # Return data}
*##==== Example of use: =====*
growth.sim(10, 4, gammas = c(300, 25),
        G = matrix(c(0.1, 0,
                     0, 0.01), nrow = 2))

         [[alternative HTML version deleted]]

      [[alternative HTML version deleted]]

        [[alternative HTML version deleted]]

	[[alternative HTML version deleted]]