simjm: Simulate data from a joint model of multiple continuous...
In moreno-betancur/survtd: Survival analysis with time-dependent covariates

Description Usage Arguments Details Value References Examples

simjm simulates data from a Cox proportional hazards or semi-parametric additive model with two time-fixed covariates (Z1 and Z2) and three time-dependent covariates (Yij_1,Yij_2,Yij_3). The user can specify various characteristics of these distributions.

1 2	simjm(n = 200, surv_model = "Cox", marker_model = "RE", MErr = "High", Miss = "Low", effects = "Weak", corr = "Low")

`n`	Size of dataset to be generated
`surv_model`	Model for the time-to-event. Options are "Cox" for Cox proportional hazard model and "Add" for semi-parametric additive hazards model.
`marker_model`	Model for the multiple markers. Options are "RE" for the correlated random effects model and "PN" for the product normal model.
`MErr`	Degree of measurement error in the multiple markers. Options are "Low", "Mod" and "High".
`Miss`	Degree of missing at-risk measurements. Options are "None", "Low" and "High".
`effects`	Strength of effects of the markers on the hazard. Options are "Null", "Weak" and "Strong".
`corr`	Degree of marginal pairwise correlations between the markers. Options are "Low", "Mod" and "High".

The function can be used to generate data from any of the scenarios considered in the the main simulation settings of Moreno-Betancur et al. (2017). See that reference for details.

A data.frame as required by survtd. That is, in the long format, with one row per individual and per visit time at which any of the time-dependent covariates were measured, with the corresponding measurements. The dataset also includes a variable that uniquely identifies observations from the same individual; a variable that indicates the timing of each measurement visit; and the fixed variables (time-to-event, event indicator, time-fixed covariates) which are constant across rows of the same individual. The final columns of the dataset (from Xij_1 onwards) are to recover the true values of the markers as per the data generation model for use with function simjm_benchmark. Specifically, the variables in the dataset are:

ID: Unique identifier of observations from the same individual.
tt: Time to event, possibly right-censored.
event: Indicator of event, with event=1 if an event occurred at tt and event=0 if the individual is censored.
Z1: Time-fixed binary covariate.
Z2: Time-fixed continuous covariate.
tj: Timing of the measurement visit.
Yij_1: Measured value of marker 1 at time tj
Yij_2: Measured value of marker 2 at time tj
Yij_3: Measured value of marker 3 at time tj
Xij_1: True value of marker 1 at time tj
Xij_2: True value of marker 2 at time tj
Xij_3: True value of marker 3 at time tj
fixed_1: Time-fixed part of the linear predictor of the linear mixed model from which Yij_1 is generated.
tim_1: Time-dependent part of the linear predictor of the linear mixed model from which Yij_1 is generated, excluding terms for other markers in the case of product-normal model.
fixed_2: Time-fixed part of the linear predictor of the linear mixed model from which Yij_2 is generated.
tim_2: Time-dependent part of the linear predictor of the linear mixed model from which Yij_2 is generated, excluding terms for other markers in the case of product-normal model.
fixed_3: Time-fixed part of the linear predictor of the linear mixed model from which Yij_3 is generated.
tim_3: Time-dependent part of the linear predictor of the linear mixed model from which Yij_3 is generated, excluding terms for other markers in the case of product-normal model.

Moreno-Betancur M, Carlin JB, Brilleman SL, Tanamas S, Peeters A, Wolfe R (2017). Survival analysis with time-dependent covariates subject to missing data or measurement error: Multiple Imputation for Joint Modeling (MIJM). Biostatistics [Epub ahead of print 12 Oct 2017].

  dat<-simjm(n=200,surv_model="Cox",marker_model="PN",
             MErr="High",Miss="None",effects="Weak",corr="Low")

  head(dat)

  dat<-simjm(n=200,surv_model="Add",marker_model="RE",
            MErr="Low",Miss="High",effects="Strong",corr="Mod")

  head(dat)