Virtual Population Data

Virtual Population Data

To get CSV files containing the Virtual Population datasets click on Generate Data in the Virtual Data section of the project. Once the data has been generated click on Download and save the resulting files.

Tip

Downloading data does not cost any additional credits and can be used to leverage the power of existing statistical or visualisation tools. This allows you to leverage the power of cloud-based computing to generate computationaly intense simulation whilst maintaining the flexibility of a modern data science toolset.

What’s included in the Virtual Population Data

A separate file is generated for each simulation and for each scenario. For example, 500 simulations with 5 scenarios will generate a total of 2500 csv files. A master file, called “MANIFEST.csv” is also produced which holds the key to each of the generated files. For the example of 500 simulations with 5 scenarios the master file will contain 2500 rows where each row contains the filename for a unique simulation and for one scenario. This information can then be merged with the file containing the simulated data and used for analysis outside of KerusCloud. To enable this, a file containing variable metadata called “METADATA.json” has been provided which details variable parameters and any advanced options and estimand strategies

Note

It is worth noting that the order in which the files appear in the “MANIFEST.csv” file may not match the order of file names created by a file list command.

How is each simulation file composed?

Each data file contains N rows per Group * Recruitment Sites where N is the Max Cohort Size for the Virtual Population. The downloaded data files contain the variable values after Advanced Options and Estimands have been applied. It should be noted that where any Imputation or Estimand replacement strategies involve calculations on the observed data (e.g. replacement with the mean of a variable) then the replacement values in the downloaded data may differ from those calculated for different sample sizes and allocations, as a result of the calculations being evaluated on different subjects. This may need to be considered if custom analysis is performed on the downloaded data.

Note

KerusCloud allows you to design various study design features, such as including multiple allocations, sites and ability to estimate the sample size.
These occur after the virtual population has been generated, so more patients may be generated than may be required to ensure an adequate number of patients exist for your desired study design.