simulate_singlecluster: Simulation of data with a censored covariate

In retogerber/censcyt: Differential abundance analysis with a right censored covariate in high-dimensional cytometry

Simulation of data with a censored covariate

Description

Function to simulate an association between a censored covariate and a predictor.

Usage

simulate_singlecluster(
  n,
  formula,
  type = c("lm", "glm", "glmer"),
  b = NULL,
  n_levels_fixeff = NULL,
  n_levels_raneff = NULL,
  weibull_params = list(X = list(shape = 0.5, scale = 0.25), C = list(shape = 1, scale
    = 0.25)),
  censoring_dependent_on_covariate = FALSE,
  weibull_params_covariate_dependent_censoring = list(shape = 1, scale = 0.1),
  error_variance = 0,
  variance_raneff = 0.5,
  transform_fn = "identity",
  verbose = FALSE
)

Arguments

n	number of samples
formula	the formula to specify the structure in the data. The censored variable should be written in the following format: 'Surv(X,I)', where 'X' is the observed value, and 'I' is the event indicator (1 if observed, 0 if censored). A full example is: 'Y ~ Surv(X,I) + Covariate + (1|Random_effect)'.
type	which regression type is used, one of 'lm', 'glm', 'glmer'. For the generalized linear models the response is binomial with a logistic link function. default = 'lm'.
b	the regression coefficients, either NUll will us 0 for the intercept and 1 for the remaining coefficients a vector with regression coefficients the length has to be (1 (intercept) + number of covariates (including the censored covariate))
n_levels_fixeff	The number of levels to use for each covariate, e.g. for two covariates: c(10,100). If NULL sets all to 2 (two groups).
n_levels_raneff	The number of levels to use for each random effect. If NULL sets to 'n' (observation level random effects).
weibull_params	The parameters for the distribution of the censored variable and the censoring time. Should be a list of lists, where the elements of the outer lists are 'X' the true value and 'C' the censoring time. The inner lists should have two keywords, 'shape' and 'scale', for the parameters of the Weibull distribution (See Weibull).
censoring_dependent_on_covariate	Logical. If censoring should depend on a covariate. The respective covariate needs to have only two levels ('n_level_fixeff'=2). Will use first covariate in formula.
weibull_params_covariate_dependent_censoring	list with two elements, shape and scale, representing the parameters of a weibull distribution for the second level of a covariate if 'censoring_dependent_on_covariate'=TRUE.
error_variance	positive double. Variance of additional gaussian noise to add in the linear sum of the predictors. For linear regression this is the only error added. Otherwise it should be set to zero. default = 0.
variance_raneff	positive double vector of the length of 'n_levels_raneff'. The variance of the gaussian distributed random effect covariates. default = 0.5.
transform_fn	function to transform censored covariate or one of 'identity' (no transformation), 'boxcox' (box-cox transformation), 'boxcox_positive' (box-cox transformation and translation to all positive values), 'log_positive' (log transformation and translation to all positive values). The transformation is applied before the response is modeled. default = 'identity'.
verbose	verbose

Value

tibble

Examples

# single differential cluster
 glmer_formula <- formula(Y ~ Surv(X,I) + Z + (1|R))
 simulate_singlecluster(100, glmer_formula, type = "glmer")
 

retogerber/censcyt documentation built on Feb. 7, 2023, 9:56 a.m.