# ‘dstidyverse’: An Implementation of  TidyverseWithin the DataSHIELD  Ecosystem

## TL;DR

The dsTidyverse package adds user-friendly data manipulation tools to the DataSHIELD platform, making it easier to analyze data without sharing individual participant information.

## Contribution

The novel contribution is implementing Tidyverse-style data manipulation functions within the DataSHIELD framework, with built-in privacy protections.

## Key findings

- dsTidyverse enables common data manipulation tasks like filtering, renaming, and grouping within DataSHIELD.
- The package includes disclosure checks to prevent data leakage while performing these operations.
- Examples show how dsTidyverse simplifies workflows for users of the DataSHIELD platform.

## Abstract

DataSHIELD is a mature, R-based federated learning platform that enables multi-site analysis without sharing individual participant data. While DataSHIELD includes many packages for data analysis, it lacks user-friendly data manipulation tools.

To address this gap, we developed
dsTidyverse, an implementation of selected functions from the popular Tidyverse package within the DataSHIELD client-server architecture. Disclosure checks were implemented to prevent individual-level data leakage.

This package provides functionality for selecting, renaming, and creating columns; conditional recoding; combining data frames by rows or columns; filtering and arranging rows; grouping and ungrouping data; and converting data frames to tibbles. Through examples, we demonstrate how
dsTidyverse simplifies common data manipulation tasks within DataSHIELD.

By providing additional data manipulation functionality,
dsTidyverse improves the user experience and analytical efficiency within DataSHIELD. The package is open-source and freely available on CRAN and GitHub, and welcomes further development:
https://github.com/molgenis/ds-tidyverse.