Generalized network-based dimensionality reduction and analysis

nda builds interpretable latent variables from communities in a squared-association network. Version 0.3.0 adds a faster and safer computation path while retaining the original numeric method codes.

Core capabilities include:

• full, partial, and semipartial Pearson, rank, robust, and distance associations;
• centrality-weighted latent variables with optional fixed communities;
• sparse networks, multicore pair calculations, FDR screening, and bootstrap loading intervals;
• printable static and interactive network plots and 2D/3D biplots;
• supervised NDRLM models using ordinary least squares, ridge, elastic net, or LOESS.

Author

• Zsolt T. Kosztyan
• Marcell T. Kurbucz
• Attila I. Katona
• Zahid Khan

Contributor

• Zsolt T. Kosztyan

Maintainer

• Zsolt T. Kosztyan

Installation

Install the development version with:

remotes::install_github("kzst/nda")
library(nda)

Quick start

fit <- ndr(swiss, cor_method = "spearman", cor_type = "full",
           centrality = "pagerank")
summary(fit)

# Assignment is silent; printing renders the graph.
network_plot <- plot(fit, interactive = FALSE, show_weights = TRUE)
network_plot

X <- as.data.frame(freeny.x)
Y <- data.frame(income = freeny.y)
model <- ndrlm(Y, X, optimize = FALSE, regression = "ridge")
predict(model, X[1:5, ])