Freedom251 › Statistical Analysis Software

partition

Free plan Linux

What partition is

partition is a free Linux desktop R framework for agglomerative partitioning. It reduces a data frame to as few variables as possible while meeting a user-specified minimum information threshold. The method is aimed at highly correlated data, including genomic datasets with redundant features. Results remain interpretable: each reduced variable maps to one or more originals, and every original maps to exactly one reduced variable. Users can choose variable selection, information-loss measurement, and the reduction process, or create custom partitioners. Built-in methods use intraclass correlation, K-means, minimum R-squared, principal component 1, and standardized mutual information. Plotting functions return ggplots and support partition visualizations through ggplot2. Permutation testing compares observed statistics with permuted data, using 100 permutations by default. super_partition() divides large feature sets into clusters before processing, with a default maximum cluster size of 4000. Data import uses R data frames or tibbles. The MIT-licensed package installs from CRAN or GitHub.

Who it suits

partition suits R users working with correlated or high-dimensional data, especially genomic features. It is useful when reducing variables while retaining an explicit information threshold and interpretable mappings.

What it does well

Free MIT-licensed package
Custom partitioners supported
Several built-in reduction methods
Permutation testing included
Handles large feature sets

What to know first

Linux desktop deployment listed
Requires R data frames or tibbles
Performance varies by partitioning method
Uses a default 4,000-feature cluster size

Verdict

partition offers configurable dimensionality reduction with mappings that remain easy to interpret. Its methods, plots, permutation tests, and large-feature handling make it suited to scripted R analysis rather than general desktop use.

Facts from partition's own pages

Authors: The site lists Joshua Millstein, Malcolm Barrett, and Katelyn Queen as developers, with Malcolm Barrett identified as maintainer. · uscbiostats.github.io, 12 Oct 2026
Built-in methods: Built-in partitioners include ICC, K-means, minimum R-squared, principal component 1, and standardized mutual information options. · uscbiostats.github.io, 12 Oct 2026
Citation: The site cites a 2020 Bioinformatics paper titled “Partition: A Surjective Mapping Approach for Dimensionality Reduction.” · uscbiostats.github.io, 11 Oct 2026
Customizable: Users can customize which variables are selected, how information loss is measured, and how data is reduced. · uscbiostats.github.io, 12 Oct 2026
Customization: Users can customize variable selection, how information loss is measured, and how data is reduced. · uscbiostats.github.io, 11 Oct 2026
Custom partitioners: Users can create custom partitioners from functions that select variables, measure information loss, and reduce data. · uscbiostats.github.io, 12 Oct 2026
Data import formats: R data frames/tibbles · uscbiostats.github.io, 23 Sept 2026
Deployment: desktop · uscbiostats.github.io, 23 Sept 2026
Developers: The site lists Joshua Millstein, Malcolm Barrett, and Katelyn Queen as developers and identifies Malcolm Barrett as maintainer. · uscbiostats.github.io, 11 Oct 2026
Free plan: Yes · uscbiostats.github.io, 23 Sept 2026
Information threshold: The partition() function reduces a data frame to as few variables as possible while meeting a user-specified information threshold. · uscbiostats.github.io, 12 Oct 2026
Installation: The package can be installed from CRAN or from the USCbiostats GitHub repository. · uscbiostats.github.io, 12 Oct 2026
Interpretability: Each original variable maps to exactly one variable in the reduced data set. · uscbiostats.github.io, 12 Oct 2026
Large feature sets: super_partition() handles data with many features by first dividing it into smaller clusters, with a default maximum cluster size of 4000. · uscbiostats.github.io, 12 Oct 2026
License: partition is released under the MIT license. · uscbiostats.github.io, 12 Oct 2026
Methods: Built-in partitioners include methods based on intraclass correlation, K-means, minimum R-squared, principal component 1, and standardized mutual information. · uscbiostats.github.io, 11 Oct 2026
Performance: The documentation reports benchmarking on a data frame with one million rows using a 2017 MacBook Pro with 16 GB RAM. · uscbiostats.github.io, 12 Oct 2026
Permutation testing: test_permutation() compares observed partition statistics with summaries from permuted data sets and defaults to 100 permutations. · uscbiostats.github.io, 12 Oct 2026
Purpose: partition is an R framework for agglomerative data reduction using a Direct-Measure-Reduce approach. · uscbiostats.github.io, 12 Oct 2026
Scaling: The site says the default ICC method scales more linearly than K-means methods as the number of columns exceeds the number of rows, while K-means is often faster at lower dimensions. · uscbiostats.github.io, 11 Oct 2026
Scripting support: Yes · uscbiostats.github.io, 23 Sept 2026
Statistical tests: Yes · uscbiostats.github.io, 23 Sept 2026
Support: The project directs users to GitHub to report bugs and provides contributor guidelines and a code of conduct. · uscbiostats.github.io, 12 Oct 2026
Use case: The package is particularly useful for highly correlated data, including genomic data with redundant features. · uscbiostats.github.io, 12 Oct 2026

Where partition stands on Freedom251

Statistical Analysis Software#73 of 106