simglm (version 0.8.0)

sim_glm: Master generalized simulation function.

Description

Takes simulation parameters as inputs and returns simulated data.

Usage

sim_glm(
  fixed,
  random,
  random3,
  fixed_param,
  random_param = list(),
  random_param3 = list(),
  cov_param,
  k,
  n,
  p,
  data_str,
  cor_vars = NULL,
  fact_vars = list(NULL),
  unbal = list(level2 = FALSE, level3 = FALSE),
  unbal_design = list(level2 = NULL, level3 = NULL),
  contrasts = NULL,
  outcome_type,
  cross_class_params = NULL,
  knot_args = list(NULL),
  ...
)

Arguments

fixed

One sided formula for fixed effects in the simulation. To suppress intercept add -1 to formula.

random

One sided formula for random effects in the simulation. Must be a subset of fixed.

random3

One sided formula for random effects at third level in the simulation. Must be a subset of fixed (and likely of random).

fixed_param

Fixed effect parameter values (i.e. beta weights). Must be same length as fixed.

random_param

A list of named elements that must contain:

• random_var = variance of random parameters,

• rand_gen = Name of simulation function for random effects.

Optional elements are:

• ther: Theorectial mean and variance from rand_gen,

• ther_sim: Simulate mean/variance for standardization purposes,

• cor_vars: Correlation between random effects,

• ...: Additional parameters needed for rand_gen function.

random_param3

A list of named elements that must contain:

• random_var = variance of random parameters,

• rand_gen = Name of simulation function for random effects.

Optional elements are:

• ther: Theorectial mean and variance from rand_gen,

• ther_sim: Simulate mean/variance for standardization purposes,

• cor_vars: Correlation between random effects,

• ...: Additional parameters needed for rand_gen function.

cov_param

List of arguments to pass to the continuous generating function, must be the same order as the variables specified in fixed. This list does not include intercept, time, factors, or interactions. Required arguments include:

• dist_fun: This is a quoted R distribution function.

• var_type: This is the level of variable to generate. Must be either 'single', 'level1', 'level2', or 'level3'. Must be same order as fixed formula above.

Optional arguments to the distribution functions are in a nested list, see the examples or vignettes for example code.

k

Number of third level clusters.

n

Cluster sample size.

p

Within cluster sample size.

data_str

Type of data. Must be "cross", "long", or "single".

cor_vars

A vector of correlations between variables.

fact_vars

A nested list of factor, categorical, or ordinal variable specification, each list must include:

• numlevels = Number of levels for ordinal or factor variables.

• var_type = Must be 'single', 'level1', 'level2', or 'level3'.

Optional arguments include:

• replace

• prob

• value.labels

See also sample for use of these optional arguments.

unbal

A named TRUE/FALSE list specifying whether unbalanced simulation design is desired. The named elements must be: "level2" or "level3" representing unbalanced simulation for level two and three respectively. Default is FALSE, indicating balanced sample sizes at both levels.

unbal_design

When unbal = TRUE, this specifies the design for unbalanced simulation in one of two ways. It can represent the minimum and maximum sample size within a cluster via a named list. This will be drawn from a random uniform distribution with min and max specified. Secondly, the actual sample sizes within each cluster can be specified. This takes the form of a vector that must have the same length as the level two or three sample size. These are specified as a named list in which level two sample size is controlled via "level2" and level three sample size is controlled via "level3".

contrasts

An optional list that specifies the contrasts to be used for factor variables (i.e. those variables with .f or .c). See contrasts for more detail.

outcome_type

A vector specifying the type of outcome, must be either logistic or poisson. Logitstic outcome will be 0/1 and poisson outcome will be counts.

cross_class_params

A list of named parameters when cross classified data structures are desired. Must include the following arguments:

• num_ids: The number of cross classified clusters. These are in addition to the typical cluster ids

• random_param: This argument is a list of arguments passed to sim_rand_eff. These must include:

  • random_var: The variance of the cross classified random effect

  • rand_gen: The random generating function used to generate the cross classified random effect.

  Optional elements are:

  • ther: Theorectial mean and variance from rand_gen,

  • ther_sim: Simulate mean/variance for standardization purposes,

  • cor_vars: Correlation between random effects,

  • ...: Additional parameters needed for rand_gen function.

knot_args

A nested list of named knot arguments. See sim_knot for more details. Arguments must include:

• var

• knot_locations

...

Not currently used.

Details

Simulated data is useful for classroom demonstrations and to study the impacts of assumption violations on parameter estimates, statistical power, or empirical type I error rates.

This function allows researchers a flexible approach to simulate regression models, including single level models and cross sectional or longitudinal linear mixed models (aka. hierarchical linear models or multilevel models).

Examples

# NOT RUN {
# generating parameters for single level regression
set.seed(2)
fixed <- ~1 + act + diff + numCourse + act:numCourse
fixed_param <- c(0.1, -0.2, 0.15, 0.5, -0.02)
cov_param <- list(dist_fun = c('rnorm', 'rnorm', 'rnorm'),
   var_type = c("single", "single", "single"),
   opts = list(list(mean = 0, sd = 4),
   list(mean = 0, sd = 3),
   list(mean = 0, sd = 3)))
n <- 150
temp_single <- sim_glm(fixed = fixed, fixed_param = fixed_param, 
  cov_param = cov_param, n = n, data_str = "single", outcome_type = 'logistic')
  
  # counts
temp_single <- sim_glm(fixed = fixed, fixed_param = fixed_param, 
  cov_param = cov_param, n = n, data_str = "single", outcome_type = 'poisson')

# Longitudinal linear mixed model example
fixed <- ~1 + time + diff + act + time:act
random <- ~1 + time + diff
fixed_param <- c(0.1, -0.2, 0.15, 0.5, -0.02)
random_param <- list(random_var = c(7, 4, 2), rand_gen = 'rnorm')
cov_param <- list(dist_fun = c('rnorm', 'rnorm'),
   var_type = c("level1", "level2"),
   opts = list(list(mean = 0, sd = 1.5),
   list(mean = 0, sd = 4)))
n <- 150
p <- 30
data_str <- "long"
temp_long <- sim_glm(fixed, random, random3 = NULL, fixed_param, 
random_param, random_param3 = NULL,
 cov_param, k = NULL, n, p, data_str = data_str, outcome_type = 'logistic')

 # counts 
temp_long <- sim_glm(fixed, random, random3 = NULL, fixed_param, 
random_param, random_param3 = NULL,
 cov_param, k = NULL, n, p, data_str = data_str, outcome_type = 'poisson')

# Three level example
fixed <- ~1 + time + diff + act + actClust + time:act
random <- ~1 + time + diff
random3 <- ~ 1 + time
fixed_param <- c(0.1, -0.2, 0.15, 0.5, -0.02, 0.03)
random_param <- list(random_var = c(7, 4, 2), rand_gen = 'rnorm')
random_param3 <- list(random_var = c(4, 2), rand_gen = 'rnorm')
cov_param <- list(dist_fun = c('rnorm', 'rnorm', 'rnorm'), 
   var_type = c("level1", "level2", "level3"),
   opts = list(list(mean = 0, sd = 1.5),
   list(mean = 0, sd = 4),
   list(mean = 0, sd = 2)))
k <- 10
n <- 15
p <- 10
data_str <- "long"
temp_three <- sim_glm(fixed, random, random3, fixed_param, random_param, 
  random_param3, cov_param, k,n, p, data_str = data_str, outcome_type = 'logistic')
  
  # count data sim
  temp_three <- sim_glm(fixed, random, random3, fixed_param, random_param, 
  random_param3, cov_param, k,n, p, data_str = data_str, outcome_type = 'poisson')


# }