fastmed

Overview

fastmed is an R package designed to perform efficient and scalable mediation analysis using multiple linear regressions and bootstrap resampling. Leveraging the power of C++ through Rcpp and parallel processing with RcppParallel, fastmed is optimized to handle large datasets and provide real-time results output to CSV files.

Features

• Efficient Mediation Analysis: Conduct mediation analysis with multiple exposure, mediator, and outcome variables.
• Parallel Processing: Utilize multiple CPU cores to accelerate computations.
• Bootstrap Resampling: Estimate direct, indirect, and total effects with confidence intervals and p-values.
• Customizable Column Prefixes: Easily specify prefixes to identify exposure, mediator, and outcome variables.
• Scalable for Large Datasets: Handle large-scale data efficiently by processing in chunks.
• Real-time CSV Output: Save results directly to CSV files during analysis.

Installation

You can install the development version of fastmed from GitHub using the remotes package:

# Install remotes if not already installed
if (!requireNamespace("remotes", quietly = TRUE)) {
  install.packages("remotes")
}

# Install fastmed from GitHub
remotes::install_github("daehwankim12/fastmed")

Usage

Here's a detailed example of how to use the mediation_analysis function in fastmed:

library(fastmed)
library(data.table)

# Generate example data
set.seed(123)
n <- 1000
my_data <- data.table(
  Exposure_A1 = rnorm(n),
  Exposure_A2 = rnorm(n),
  Mediator_X1 = rnorm(n),
  Mediator_X2 = rnorm(n),
  Outcome_Y1 = rnorm(n),
  Outcome_Y2 = rnorm(n)
)

# Define column prefixes
columns <- list(
  exposure = "Exposure_A",
  mediator = "Mediator_X",
  outcome = "Outcome_Y"
)

# Specify output CSV file path
output_csv <- "mediation_results.csv"

# Perform mediation analysis
mediation_analysis(
  data = my_data,
  columns = columns,
  nrep = 500,            # Number of bootstrap replicates
  output_file = output_csv,
  num_threads = 4        # Number of threads for parallel processing
)

# View results
results <- fread(output_csv)
print(results)

Parameters

• data: A data.table or data.frame containing the dataset.
• columns: A list with three named elements: exposure, mediator, and outcome. Each should be a character string specifying the prefix of the respective columns.
• nrep: (Optional) Number of bootstrap replicates. Default is 1000.
• output_file: Path to the output CSV file where results will be saved.
• num_threads: (Optional) Number of threads for parallel processing. Defaults to the number of available cores.

Output

The output CSV file will contain detailed results for each combination of exposure, mediator, and outcome variables, including estimates, standard errors, confidence intervals, and p-values for indirect, direct, and total effects.

Performance Considerations

• The package is optimized for parallel processing. Increase num_threads to utilize more CPU cores.
• For very large datasets, consider splitting the analysis into smaller chunks and combining the results.
• Monitor memory usage, especially when increasing nrep for bootstrap resampling.

Troubleshooting

If you encounter issues:

1. Ensure you have the latest version of fastmed installed.
2. Check that all dependencies are up to date.
3. For performance issues, try adjusting num_threads or reducing nrep.
4. If you encounter a bug, please open an issue with a reproducible example.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

fastmed builds upon several powerful R packages:

• Rcpp for C++ integration
• RcppParallel for parallel processing
• data.table for efficient data manipulation

Citation

If you use fastmed in your research, please cite it as follows:

Kim, D. (2024). fastmed: Fast Mediation Analysis in R. R package version 0.1.0.
https://github.com/daehwankim12/fastmed

Contact

For questions, suggestions, or collaborations, please open an issue on the GitHub repository or contact the package maintainer at [email protected].