Functions for Multi-Dimensional Analysis

Multi-Dimensional Analysis (MDA) is an adaptation of factor analysis developed by Douglas Biber (1992) . Its most common use is to describe language as it varies by genre, register, and use. This package contains functions for carrying out the calculations needed to describe and plot MDA results: dimension scores, dimension means, and factor loadings.

mda.biber

An R package for conducting Multi-Dimensional Analysis (MDA), a statistical procedure developed by Douglas Biber for analyzing linguistic variation across genres, registers, and text types.

About Multi-Dimensional Analysis

Multi-Dimensional Analysis (MDA) is a specialized application of factor analysis that identifies co-occurring patterns in linguistic features. It is based on the principle that some linguistic variables co-occur (e.g., nouns and adjectives) while others inversely co-occur (e.g., nouns and pronouns).

MDA typically involves:

1. Feature extraction: Measuring syntactic categories (passive constructions), lexical classes (nouns), or functional categories (hedges)
2. Factor analysis: Identifying underlying dimensions of linguistic variation
3. Dimension scoring: Computing scores for texts along each dimension
4. Interpretation: Understanding what linguistic patterns each dimension represents

The linguistic features used in this package are based on Biber's original research, documented in the corpora and pseudobibeR packages.

Key Features

• Core MDA function: mda_loadings() performs factor analysis and calculates dimension scores
• Visualization tools: Multiple plotting functions for different aspects of MDA results
• Built-in dataset: MICUSP corpus data tagged with 67 linguistic features
• Flexible parameters: Customizable correlation thresholds and factor loadings
• Comprehensive output: Dimension scores, group means, and factor loadings

Installation

Install the development version from GitHub:

# Install from GitHub
devtools::install_github("browndw/mda.biber")

# Load the package
library(mda.biber)

Quick Start

library(mda.biber)
library(dplyr)

# Load and prepare the built-in MICUSP dataset
data(micusp_biber)

# Extract discipline codes and convert to factor
d <- micusp_biber %>%
  mutate(doc_id = stringr::str_extract(doc_id, "^[A-Z]+")) %>%
  mutate(doc_id = as.factor(doc_id)) %>%
  select(where(~ any(. != 0)))  # Remove zero-only columns

# Determine optimal number of factors
screeplot_mda(d)

# Perform MDA with 2 factors
mda_result <- mda_loadings(d, n_factors = 2)

# View dimension scores
head(mda_result)

# Access group means and factor loadings
attributes(mda_result)$group_means
attributes(mda_result)$loadings

Visualization

The package provides several visualization functions:

Stick Plot

# Show category means along a dimension
stickplot_mda(mda_result, n_factor = 1)

Heatmap

# Combine stick plot with factor loadings heatmap
heatmap_mda(mda_result, n_factor = 1)

Boxplot

# Show dimension scores with contributing variables
boxplot_mda(mda_result, n_factor = 1)

Dataset

The package includes micusp_biber, a dataset containing:

• 828 texts from the Michigan Corpus of Upper-Level Student Papers (MICUSP)
• 67 linguistic features tagged using the pseudobibeR package
• Multiple disciplines represented (Biology, Engineering, Philosophy, etc.)
• Standardized rates per 1,000 tokens for most features

Functions Overview

Documentation

• Package documentation
• Full methodology description: Multi-dimensional analysis tutorial

Citation

If you use this package in your research, please cite:

Brown, D. (2024). mda.biber: Functions for Multi-Dimensional Analysis. 
R package version 1.0.1. https://github.com/browndw/mda.biber

License

This package is licensed under the MIT License.