Simulate a "clinical" dataset according to assumptions of the MMRM. Datasets should have "observations" at a selected number of discrete timepoints. Observations are predicted by:
the mean observation at each timepoint
the effect of baseline at each timepoint
the effect of arm at each timepoint
residual covariance matrix
Usage
mmrm_simulate(
n_per_arm = 50,
n_timepoints = 5,
effect_tp = 8,
effect_baseline = -0.5,
effect_arm = 2,
covariance = NULL,
variance = 10,
correlation = 0,
baseline = c(25, 5),
p_missing = 0.1
)Arguments
- n_per_arm
the number of subjects per arm
- n_timepoints
the number of timepoints. If effects are specified as vectors, this is ignored.
- effect_tp
mean observation at each timepoint. Either a vector with length equal to the number of timepoints or a scalar.
- effect_baseline
effect of baseline value at each timepoint. Either a vector with length equal to the number of timepoints or a scalar.
- effect_arm
effect of arm at each timepoint. Either a vector with length equal to the number of timepoints or a scalar.
- covariance
full residual covariance matrix. Either covariance of both of variance and correlation should be specified.
- variance
variance at each timepoint. Either a vector with length equal to the number of timepoints or a scalar. Either both of variance and correlation or covariance should be specified
- correlation
correlation of residuals across timepoints. Should be a scalar or vector with length: timepoints * (timepoints - 1) / 2. Either both of variance and correlation or covariance should be specified.
- baseline
mean and variance of baseline scores
- p_missing
probability of missing data at each timepoint. Either a vector with length equal to the number of timepoints or a scalar.