Rmcp

What is Rmcp

rmcp is an R Econometrics MCP Server that provides econometric modeling capabilities through R, enabling AI assistants to perform advanced econometric analyses such as linear regression and panel data models.

Use cases

Use cases include analyzing the relationship between variables in datasets, conducting panel data studies, estimating causal effects using instrumental variables, and performing diagnostic tests on econometric models.

How to use

To use rmcp, install the required Python and R packages, then run the server either via Docker or manually. Connect it to Claude Desktop by adding a new server configuration.

Key features

Key features include linear regression with robust standard errors, panel data analysis (fixed effects, random effects), instrumental variables regression, diagnostic tests for econometric models, and pre-defined prompt templates for common analyses.

Where to use

rmcp can be used in fields such as economics, finance, and data science where econometric analysis is essential for understanding relationships between variables.

R MCP Server

A Model Context Protocol (MCP) server that provides advanced econometric modeling and data analysis capabilities through R. This server enables AI assistants to perform sophisticated econometric and statistical analyses seamlessly, helping you quickly gain insights from your data.

Features

• Linear Regression: Run linear models with optional robust standard errors.
• Panel Data Analysis: Estimate fixed effects, random effects, pooling, between, and first-difference models.
• Instrumental Variables: Build and estimate IV regression models.
• Diagnostic Tests: Assess heteroskedasticity, autocorrelation, and model misspecification.
• Descriptive Statistics: Generate summary statistics for datasets using R’s summary() functionality.
• Correlation Analysis: Compute Pearson or Spearman correlations between variables.
• Group-By Aggregations: Group data by specified columns and compute summary statistics using dplyr.
• Resources: Access reference documentation for various econometric techniques.
• Prompts: Use pre-defined prompt templates for common econometric analyses.

Installation

Using Docker (Recommended)

1. Build the Docker image:

   docker build -t r-econometrics-mcp .
   

2. Run the container:

docker run -it r-econometrics-mcp

Manual Installation

Install the required Python packages:

pip install -r requirements.txt

Install the required R packages (if you run the server outside a container):

install.packages(c("plm", "lmtest", "sandwich", "AER", "jsonlite"), repos="https://cloud.r-project.org/")

Run the server:

python rmcp.py

Usage

The server communicates via standard input/output. When you run:

python rmcp.py

it starts and waits for JSON messages on standard input. To test the server manually, create a file (for example, test_request.json) with a compact (single-line) JSON message.

Example Test

Create test_request.json with the following content (a one-line JSON):

{
  "tool": "linear_model",
  "args": {
    "formula": "y ~ x1",
    "data": {
      "x1": [
        1,
        2,
        3,
        4,
        5
      ],
      "y": [
        1,
        3,
        5,
        7,
        9
      ]
    },
    "robust": false
  }
}

Then run:

cat test_request.json | python rmcp.py

Output

{"coefficients": {"(Intercept)": -1, "x1": 2}, "std_errors": {"(Intercept)": 2.8408e-16, "x1": 8.5654e-17}, "t_values": {"(Intercept)": -3520120717017444, "x1": 23349839270207356}, "p_values": {"(Intercept)": 5.0559e-47, "x1": 1.7323e-49}, "r_squared": 1, "adj_r_squared": 1, "sigma": 2.7086e-16, "df": [2, 3, 2], "model_call": "lm(formula = formula, data = data)", "robust": false}

Usage with Claude Desktop

1. Launch Claude Desktop
2. Open the MCP Servers panel
3. Add a new server with the following configuration:
   • Name: R Econometrics
   • Transport: stdio
   • Command: path/to/python r_econometrics_mcp.py
   • (Or if using Docker): docker run -i r-econometrics-mcp

Example Queries

Here are some example queries you can use with Claude once the server is connected:

Linear Regression

Can you analyze the relationship between price and mpg in the mtcars dataset using linear regression?

Panel Data Analysis

I have panel data with variables gdp, investment, and trade for 30 countries over 20 years. Can you help me determine if a fixed effects or random effects model is more appropriate?

Instrumental Variables

I'm trying to estimate the causal effect of education on wages, but I'm concerned about endogeneity. Can you help me set up an instrumental variables regression?

Diagnostic Tests

After running my regression model, I'm concerned about heteroskedasticity. Can you run appropriate diagnostic tests and suggest corrections if needed?

Tools Reference

linear_model

Run a linear regression model.

Parameters:

• formula (string): The regression formula (e.g., ‘y ~ x1 + x2’)
• data (object): Dataset as a dictionary/JSON object
• robust (boolean, optional): Whether to use robust standard errors

panel_model

Run a panel data model.

Parameters:

• formula (string): The regression formula (e.g., ‘y ~ x1 + x2’)
• data (object): Dataset as a dictionary/JSON object
• index (array): Panel index variables (e.g., [‘individual’, ‘time’])
• effect (string, optional): Type of effects: ‘individual’, ‘time’, or ‘twoways’
• model (string, optional): Model type: ‘within’, ‘random’, ‘pooling’, ‘between’, or ‘fd’

diagnostics

Perform model diagnostics.

Parameters:

• formula (string): The regression formula (e.g., ‘y ~ x1 + x2’)
• data (object): Dataset as a dictionary/JSON object
• tests (array): Tests to run (e.g., [‘bp’, ‘reset’, ‘dw’])

iv_regression

Estimate instrumental variables regression.

Parameters:

• formula (string): The regression formula (e.g., ‘y ~ x1 + x2 | z1 + z2’)
• data (object): Dataset as a dictionary/JSON object

Resources

• econometrics:formulas: Information about common econometric model formulations
• econometrics:diagnostics: Reference for diagnostic tests
• econometrics:panel_data: Guide to panel data analysis in R

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License