software.Rmd

---
title: "Software"
output:
  html_document:
    includes:
       in_header: analytics.html
---

<link rel="stylesheet" href="academicons.css"/>

<br>

### SOFTWARE

A selection of R packages I have developed. For other code and analysis scripts, see my [GitHub profile](http://github.com/angelgar/).

***

#### afpca — Adaptive Functional Principal Component Analysis

An R package for estimating directions of variation in functional data that exhibit sharp changes in smoothness. `afpca` combines a fast, scalable adaptive scatterplot smoother with a probabilistic FPCA framework, allowing each functional principal component to be smoothed adaptively. This is particularly useful in settings such as neural recordings, where sharp post-stimulus transitions must be distinguished from smooth baseline behavior — settings where standard global-smoothness assumptions fail.

* Source: [github.com/angelgar/afpca](https://github.com/angelgar/afpca)
* Documentation & vignettes: [angelgarciadelagarza.com/afpca](http://angelgarciadelagarza.com/afpca/)
* Co-developed with [Britton Sauerbrei](https://sauerbreilab.org/) and [Jeff Goldsmith](https://jeffgoldsmith.com/).
* Companion manuscript: [Adaptive Smoothing Functional Principal Component Analysis](https://arxiv.org/abs/2310.01760) (under review at *Biometrics*).

Install the development version from GitHub:

```r
# install.packages("devtools")
devtools::install_github("angelgar/afpca")
```

**A quick visual tour.** Below are example outputs from the package vignette. First, 20 simulated curves drawn from a mean function and two functional principal components with locally varying smoothness:

<p style="text-align:center;">
<img src="https://github.com/angelgar/afpca/raw/master/man/figures/README-simulated_data_plot-1.png" alt="Simulated functional data with locally varying smoothness" style="width:100%; max-width:700px;"/>
</p>

The estimated mean function and the first two adaptively-smoothed functional principal components recovered by `fpca.adapt()` — note how the estimator preserves sharp post-onset transitions rather than over-smoothing them:

<p style="text-align:center;">
<img src="https://github.com/angelgar/afpca/raw/master/man/figures/README-estimated_fpc_plot-1.png" alt="Estimated mean function and adaptively-smoothed functional principal components" style="width:100%; max-width:700px;"/>
</p>

And two examples of observed curves alongside their reconstructions from the estimated components:

<p style="text-align:center;">
<img src="https://github.com/angelgar/afpca/raw/master/man/figures/README-reconstructed_plots-1.png" alt="Observed curves with adaptive FPCA reconstructions" style="width:100%; max-width:700px;"/>
</p>

***

#### voxel — Mass-Univariate Voxelwise Analysis of Imaging Data

An R package for mass-univariate voxelwise analysis of NIfTI medical-imaging data, supporting general linear and generalized additive model workflows at the voxel level.

* CRAN: [voxel](https://cran.r-project.org/web/packages/voxel/) — 35,000+ downloads
* Co-developed with [Simon Vandekar](https://www.vumc.org/biostatistics/person/simon-vandekar-phd), [Ted Satterthwaite](https://www.satterthwaitelab.com/), and [Taki Shinohara](https://www.pennmedicine.org/cancer/about/clinical-trials/cancer-clinical-research-team/biostatistics/taki-shinohara).

Install from CRAN:

```r
install.packages("voxel")
```

<p style="text-align:center;">
<img src="images/figure-voxel.pdf" alt="voxel package example output" style="width:100%; max-width:700px;"/>
</p>

<br>