PatientSimulation

Simulation of virtual patients

In this section will be showed how use the template to simulate a new pop- ulation of patients. 
In order to do that, the only requirement is to have a dataset containing the covariates matrix. 
The template include two different methods to simulate covariates

Resampling method - This approach consist in a random resempling of the provided dataset. 
                    In order to perform the resampling the function in the template recall the function 
                    createExternalCovariates in the package MSToolkit.

Distribution based method - This approach consist in computing a multivariate normal distribution. 
                            In order to perform the covariate simulation step, the high-level function multivariatePatients 
                            recall the low-level function function covMultivarDist.

Although the package contains specific functions for the simulation of covariates 
(i.e. createExternalCovariates from MSToolkit and covMultivarDist) the global process of patient and study design 
simulation may also be performed by high level functions (i.e. resamplingPatients and multivariatePatients). 
This high level functions recall the low level functions to simulate the covariate matrix, 
but in addition to that they also simulate the study design and allocate each patient to a specific treatment group. 
Examples of the use of these functions is provided below.


Requirements

A dataset containing the covariates, this dataset must also contain a column indicating the subject ID. 
The only mandatory columns in the dataset are the ones containing the covariates and a column containing a subject identifier; 
all the other column are ignored.

Use of resampling method

This simulation is performed using the function resamplingPatients, for more technical information on this function please 
read the function help file, available with the command ?resamplingPatients.

Example of R code

```

d <- read.csv("CS1_IV1EST_PAR.csv",header=TRUE)

head(d)

In order to make the code cleaner, we will define in R a variable for each argument of the function

Number of subject to include in the trial
This will be the total number of subjects and will be equally divided between the treatments

myPatients <- 100

Vector of doses to include in the trial.
In the real data we have 100 and 250

myDoses <- c(0,100,200)

Vector of times

myTimes <- c(0,0.2,0.5,1,2,4,6,12,15,18,24)

Variable that define the study design.

May take values "Parallel" or "Crossover"

myStudyType <- "Parallel"

Variable that define the header of the column in the dataset

reporting the subject ID

myHeader = "CID"

Vector of names indicating the covariates to sample

myCovariatesNames <- c("AGE", "WT", "SCR", "ISM", "CLCR")

Name of the dataset containing the covariates

myDataset <- "CS1_IV1EST_PAR.csv"

Path indicating where the dataset with the covariates is.

the default value is the working directory,

so if the dataset is in this directory

you can can ignore this argument

covariatePath <- getwd()

Variable that define if you want to create a new dataset

If this variable is FALSE, the function create a dataframe

In this case, you may want to use expressions

like d <- resamplingPatients(...)

to assign the dataframe to an R object

writeFile <- TRUE

Call to the function with the variables that we defined

resamplingPatients(myPatients, myDoses, myTimes, myStudyType, myHeader, myCovariatesNames, myDataset, covariatePath, writeFile)

```

Use of multivariate method

In this section it will showed how to simulate a dataset using a multivariate normal distribution to simulate the covariates. This simulation is performed with the function multivariatePatients. This function recall the package function covMultivarDist in the actual covariate simulation step. The way how to use this function is almost identical to the function resamplingPatients, the only difference is that in this case the names of the covariates to sample mast be provided dividing the continuous and the categorical ones.

Example of R code

```

d <- read.csv("CS1_IV1EST_PAR.csv",header=TRUE)

head(d)

In order to make the code cleaner, we will define in R a

variable for each argument of the function

Number of subject to include in the trial

This will be the total number of subjects

and will be equally divided between the treatments

myPatients <- 100

Vector of doses to include in the trial.

In the real data we have 100 and 250

myDoses <- c(0,100,200)

Vector of times

myTimes <- c(0,0.2,0.5,1,2,4,6,12,15,18,24)

Variable that define the study design.

May be "Parallel" or "Crossover"

myStudyType <- "Parallel"

Variable that define the header of the column in the dataset

reporting the subject ID

myHeader = "CID"

This options represent the only difference in syntax

with the function resamplingPatients()

Instead of a single vector of names containing all the

covariates to sample in this case must be provided 2 vectors,

one for the continuous covariates and one for the categoricals

catNames <- c("ISM") contNames <- c("AGE", "WT", "SCR", "CLCR")

Name of the dataset containing the covariates

myDataset <- "CS1_IV1EST_PAR.csv"

Path indicating where the dataset with the covariates is.

the default value is the working directory,

so if the dataset is in this directory

you can can ignore this argument

covariatePath <- getwd()

Variable that define if you want to create a new dataset

If this variable is FALSE, the function create a dataframe

In this case, you may want to use expressions like

d <- multivariatePatients(...) to assign the dataframe to an R object

writeFile <- TRUE

Variable describing the name of the new dataset that will be created

This variable is necessary only if writeFile=TRUE

newDatasetName <- "myTrial.csv"

We recall the function with the variables thet we defined

multivariatePatients(myPatients, myDoses, myTimes, myStudyType, myHeader, catNames, contNames, myDataset, covariatePath, writeFile, newDatasetName)