modgo_survival {modgo} | R Documentation |
MOck Data GeneratiOn
Description
modgo_survival
Create mock dataset from a real one by using
Generalized Lambdas Distributions and by seperating the data set in 2 based
in the event status.
Usage
modgo_survival(
data,
event_variable = NULL,
time_variable = NULL,
surv_method = 1,
ties_method = "max",
variables = colnames(data),
bin_variables = NULL,
categ_variables = NULL,
count_variables = NULL,
n_samples = nrow(data),
sigma = NULL,
nrep = 100,
noise_mu = FALSE,
pertr_vec = NULL,
change_cov = NULL,
change_amount = 0,
seed = 1,
thresh_var = NULL,
thresh_force = FALSE,
var_prop = NULL,
var_infl = NULL,
infl_cov_stable = FALSE,
tol = 1e-06,
stop_sim = FALSE,
new_mean_sd = NULL,
multi_sugg_prop = NULL,
generalized_mode = TRUE,
generalized_mode_model = NULL,
generalized_mode_model_event = "rprs",
generalized_mode_model_no_event = "rprs",
generalized_mode_lmbds = NULL
)
Arguments
data |
a data frame containing the data whose characteristics are to be mimicked during the data simulation. |
event_variable |
a character string listing the event variable. |
time_variable |
a character string listing the time variable. |
surv_method |
A numeric value that indicates which one of the 2 survival methods will be used. First method(surv_method = 1): Event and no event data sets are using different covariance matrices for the simulation. Second method(surv_method = 2): Event and no event data sets are using the same covariance matrix for the simulation |
ties_method |
Method on how to deal with equal values
during rank transformation. Acceptable input:"max","average","min". This
parameter is passed by |
variables |
a vector of which variables you want to transform. Default:colnames(data) |
bin_variables |
a character vector listing the binary variables. |
categ_variables |
a character vector listing the ordinal categorical variables. |
count_variables |
a character vector listing the count as a sub sub category of categorical variables. Count variables should be part of categorical variables vector. Count variables are treated differently when using gldex to simulate them. |
n_samples |
Number of rows of each simulated data set. Default is
the number of rows of |
sigma |
a covariance matrix of NxN (N= number of variables) provided by the user to bypass the covariance matrix calculations |
nrep |
number of repetitions. |
noise_mu |
Logical value if you want to apply noise to multivariate mean. Default: FALSE |
pertr_vec |
A named vector.Vector's names are the continuous variables that the user want to perturb. Variance of simulated data set mimic original data's variance. |
change_cov |
change the covariance of a specific pair of variables. |
change_amount |
the amount of change in the covariance of a specific pair of variables. |
seed |
A numeric value specifying the random seed. If |
thresh_var |
A data frame that contains the thresholds(left and right) of specified variables (1st column: variable names, 2nd column: Left thresholds, 3rd column: Right thresholds) |
thresh_force |
A logical value indicating if you want to force threshold in case the proportion of samples that can surpass the threshold are less than 10% |
var_prop |
A named vector that provides a proportion of value=1 for a specific binary variable(=name of the vector) that will be the proportion of this value in the simulated data sets.[this may increase execution time drastically] |
var_infl |
A named vector.Vector's names are the continuous variables that the user want to perturb and increase their variance |
infl_cov_stable |
Logical value. If TRUE,perturbation is applied to original data set and simulations values mimic the perturbed original data set.Covariance matrix used for simulation = original data's correlations. If FALSE, perturbation is applied to the simulated data sets. |
tol |
A numeric value that set up tolerance(relative to largest variance) for numerical lack of positive-definiteness in Sigma |
stop_sim |
A logical value indicating if the analysis should stop before simulation and produce only the correlation matrix |
new_mean_sd |
A matrix that contains two columns named "Mean" and "SD" that the user specifies desired Means and Standard Deviations in the simulated data sets for specific continues variables. The variables must be declared as ROWNAMES in the matrix |
multi_sugg_prop |
A named vector that provides a proportion of value=1 for specific binary variables(=name of the vector) that will be the close to the proportion of this value in the simulated data sets. |
generalized_mode |
A logical value indicating if generalized lambda/poisson distributions or set up thresholds will be used to generate the simulated values |
generalized_mode_model |
A matrix that contains two columns named "Variable" and "Model". This matrix can be used only if a generalized_mode_model argument is provided. It specifies what model should be used for each Variable. Model values should be "rmfmkl", "rprs", "star" or a combination of them, e.g. "rmfmkl-rprs" or "star-star", in case the use wants a bimodal simulation. The user can select Generalised Poisson model for poisson variables, but this model cannot be included in bimodal simulation |
generalized_mode_model_event |
A matrix that contains two columns named "Variable" and "Model" and it is to be used for the event data set(event = 1). This matrix can be used only if a generalized_mode_model argument is provided. It specifies what model should be used for each Variable. Model values should be "rmfmkl", "rprs", "star" or a combination of them, e.g. "rmfmkl-rprs" or "star-star", in case the use wants a bimodal simulation. The user can select Generalised Poisson model for poisson variables, but this model cannot be included in bimodal simulation |
generalized_mode_model_no_event |
A matrix that contains two columns named "Variable" and "Model" and it is to be used for the non-event data set(event = 0). This matrix can be used only if a generalized_mode_model argument is provided. It specifies what model should be used for each Variable. Model values should be "rmfmkl", "rprs", "star" or a combination of them, e.g. "rmfmkl-rprs" or "star-star", in case the use wants a bimodal simulation. The user can select Generalised Poisson model for poisson variables, but this model cannot be included in bimodal simulation |
generalized_mode_lmbds |
A matrix that contains lambdas values for each of the variables of the data set to be used for either Generalized Lambda Distribution Generalized Poisson Distribution or setting up thresholds |
Details
Simulated data is generated based on available data. The simulated data mimics the characteristics of the original data. The algorithm used is based on the ranked based inverse normal transformation (Koliopanos et al. (2023)).
Value
A list with the following components:
simulated_data |
A list of data frames containing the simulated data. |
original_data |
A data frame with the input data. |
correlations |
a list of correlation matrices. The ith element is the
correlation matrix for the ith simulated dataset. The |
bin_variables |
character vector listing the binary variables |
categ_variables |
a character vector listing the ordinal categorical variables |
covariance_matrix |
Covariance matrix used when generating observations from a multivariate normal distribution. |
seed |
Random seed used. |
samples_produced |
Number of rows of each simulated dataset. |
sim_dataset_number |
Number of simulated datasets produced. |
Author(s)
Francisco M. Ojeda, George Koliopanos
Examples
data("cancer", package = "survival")
cancer_data <- na.omit(cancer)
cancer_data$sex <- cancer_data$sex - 1
cancer_data$status <- cancer_data$status - 1
test_surv <- modgo_survival(data = cancer_data,
surv_method = 1,
bin_variables = c("status", "sex"),
categ_variables = "ph.ecog",
event_variable = "status",
time_variable = "time",
generalized_mode_model_no_event = "rmfmkl",
generalized_mode_model_event = "rprs")