Is bllflow for you?

  • Do you shudder at the thought of trying to update the analyses for a previous study? (let alone imagine someone else trying to replicate your analyses?)

  • Are your data and statistical models becoming more complex, challenging to perform, and challenging to report?

  • Are you concerned about the misuse of statistical findings? But not sure about prespecifying your model and reporting all results of all analyses?

  • Do you work in teams that span disciplines and institutions?

We answered ‘yes’ to all these questions and then created bllflow.

The purpose of bllflow

bllFlow supports transparent, reproducible data analyses and model development. The goal is to improved science quality with quicker and more efficient data analyses.

What does bllflow do?

The focus of bllflow is data cleaning and variable transformation – the most time consuming and tedious analytic task – and analyses reporting.

bllflow functions and workflow build from other packages including sjmisc, tableone, codebook, and Hmisc.

There are three main features:

  1. The Model Specification Workbook (MSW) - Start your model development with worksheets (CSV files) that contain information about the variables in your model, data cleaning and transformation steps and how to create output tables.

  2. Functions to perform routine data cleaning and transformation tasks - use functions with or without the Model Specification Workbook. Complex non-linear transformations and interactions are as easy to perform as simpler transformations. Functions with ‘BLL’ in the function name perform data cleaning and transformation using the Model Specification Workbook.

  3. Formatted output files, tables - results of your analyses in a consistent format following the concept of ‘one document, many uses’.

At any point of your analyses you have:

  • a log of data cleaning and transformed variables (how your data was cleaned and transformed).
  • a codebook to facilitate data transparency and provenance.

bllflow supports the use of metadata, including:

  • the Data Document Initiative (DDI).
  • Predictive Model Modelling Language for predictive algorithms (PMML) files for transparent algorithm reporting and deployment.

bllflow workflow and functions support reporting guidelines such as TRIPOD, STROBE, and RECORD.


# If not installed, install the devtools

# then, install the package

There are plans to submit bllFlow to CRAN once we include all seven steps of the bllflow workflow. Currently on step #4.

Contributing to the package

Please follow this guide if you like to contribute to the bllflow package.


In case your documentation .Rmd file contains images you wish to include, make sure to add the path to the image in the resource_files section. An example of that is shown below:

  - ../man/figures/coding.png

In case you wish to link .Rmd files together make sure to change their .Rmd extension to a .html extension to make sure the wiki links properly.