Create a set of time-varying covariates

Creates a set of time-varying covariates from a multivariate normal distribution and (optionally) a set of constraints.

Usage

createTimeVaryingCovariates(
  subjects,
  names,
  mean,
  covariance = 0,
  range = NULL,
  digits,
  maxDraws = 100,
  seed = .deriveFromMasterSeed(),
  idCol = getEctdColName("Subject"),
  timeCol = getEctdColName("Time"),
  treatPeriod
)

Arguments

subjects

(Required) Subjects for which to create covariates

names

(Required) Names for the continuous covariates. They should be valid R names (See validNames) and no duplicate name should be given

mean

(Required) List of means. Must be of same length than names. If it is a vector, all the time point will use the same value

covariance

(Optional) Lower triangle of covariance matrix. See parseCovMatrix for details. Should be a list. If it is a vector, all the time point will use the same value

range

(Optional) Ranges of acceptable values for each covariates. See parseRangeCode for details. It should be a list. If some covariates don't have the range, the components should be set to NULL

digits

(Optional) Number of digits used to round the values. This argument can be either missing (the default), so no rounding is done, of length one and all variables are rounded at the same digits, of same length than the number of covariates so that each covariate is rounded according to its value. This argument is first parsed by parseCharInput so it can either be a character vector or a numeric vector. See parseCharInput. If the parsed digits vector does not have length one or length equal to the number of covariates, an error is generated by the ectdStop function. This is missing by default, resulting in no rounding being performed

maxDraws

(Optional) Maximum number of attempts allowed if initial data not in range (100 by default)

seed

(Optional) Random seed to use. By default, this is derived from the current random seed

idCol

(Optional) Name of the subject column. Must be a valid R name (See validNames) and not equal to one entry of names. "SUBJ" by default

timeCol

(Optional) Name of the time column. Must be a valid R name (See validNames) and not equal to one entry of names. "TIME" by default

treatPeriod

Vector of time points

See also

createDiscreteCovariates to create covariates for a discrete distribution.

createExternalCovariates to create covariates by sampling data from an external file.

createCovariates that wraps createContinuousCovariates and the two other described above.

Author

Mike K Smith mstoolkit@googlemail.com

Examples

 # same
 dat <- createTimeVaryingCovariates(10, "X, Y, Z",
  mean <- list(X = 1:4, Y = rep(3, 4), Z = "2.5, 3, 3.2, 3.6"),
  covariance = list(1, 2:5, cbind(c(1,.5,.3,0), c(.5,1,0,0), c(.3,0,1,0), c(0,0,0,1))),
  range = list("10>=X>0", NULL, c("Z>0", "Z<=10")),
  idCol = "SUBJ", timeCol = "TIME", treatPeriod = c(0.25, 0.5, 1, 12))