🤝 Support

Contributions, issues, feature requests and sponsorship are all welcome!

Give a ⭐️ if you like this project!

Flexible Robust Statistics Data Analysis

FSDA release  2025b is out. (December 2025)

| New Features and Changes | | --- | | release_notes (HTML file) | <a href="https://youtu.be/wcVNpKE_O148"> Release notes (YouTube video) <img src="https://upload.wikimedia.org/wikipedia/commons/b/b8/YouTube_Logo_2017.svg" alt="YouTube Logo" width="100"></a>| <br>

In order to run the new features run the file below and enjoy!!!

| FileName | View :eyes:| Run ▶️ | Jupiter notebook | | -------- | ---- | --- | ---- | |New_features_FSDA2025b: examples with the new features | | | New_features_FSDA2025b.ipynb |

Running Examples on MATLAB Online

Get started with some example scripts right away using MATLAB Online. You can view or run each of the examples listed below. Sample data are downloaded when executing the scripts.

FSDA has a series of functions which complement those of the Statistics and Machine Learning toolbox.

Exploratory data analysis

| Name | Analysis Type | View :eyes: | Run ▶️| | --- | --- | --- | --- | | Missing data analysis. Discover any structure of missing observations in the data and produces a report about lower and upper univariate outliers. | Call to function mdpattern | | |
Compare robust and non robust indexes. Create histograms with grouping variable and clickable legend. | Call to functions grpstatsFS, histFS and clickableMultiLegend | | | | Label the outliers in the boxplots | Call to function add2boxplot| | |

---

Interactive Principal component analysis non robust/robust

| Name | Analysis Type | View :eyes: | Run ▶️| | --- | --- | --- | --- | |Automatically show the plots of variance explained, correlations with PCA and outlier map to find and produce a GUI written with App designer to show in an interactive way different types of biplots. <BR> Row and column points associated with arrows can be hidden or shown. The sign of the PCs can be interactively changed. The points can be shown with a color which is proportial to the othogonal distance to the space of the first 2 PCs. This enables us to immediately understand which are the units that are not well represented in the subspace formed by the first two PCs | Call to function pcaFS | | | |Interactive brushing in the space of the first two PCs. <BR> It is possible to brush a region in the biplot of the first two PCs and see the units shown in the original scatter plot matrix. Moreover if the units are are geographical coordinates and the latitude and longitude is given the geobubble plot is automatically shown. | Call to function biplotFS | | | Robust principal component analysis. <BR> It is possible to use different robust methods to find a subset of clean units. For examples both the use of MCD with a level of trimming set by the user or the forward search fixing the proportion of units to use or to have an automatic outlier detection procedure. | Call to function pcaFS with option robust set to true. | | |

Interactive Correspondence analysis non robust/robust

| Name | Analysis Type | View :eyes: | Run ▶️| | --- | --- | --- | --- | | Correspondence analysis (traditional and robust). <BR> It is possible to automatically obtain the, singular values, the inertia, explained, and cumulative. For Row and Column Points we automatically show, for each dimension: the scores Scores, th