R Econometrics MCP Server

Shared Examples =============== Cross-reference examples showing both Python package and HTTP server usage for common statistical analysis tasks. Business Analytics Example --------------------------- Analyze marketing ROI using both interfaces. Python Package ~~~~~~~~~~~~~~ .. code-block:: python # Using RMCP Python package import rmcp # Data data = { "sales": [100, 120, 115, 140, 135, 150], "marketing": [5, 8, 6, 10, 9, 12], "season": ["Q1", "Q2", "Q1", "Q2", "Q1", "Q2"] } # Linear regression result = rmcp.linear_model( formula="sales ~ marketing + season", data=data ) print(result) HTTP Server ~~~~~~~~~~~ .. code-block:: bash # Using RMCP HTTP server curl -X POST https://rmcp-server-394229601724.us-central1.run.app/mcp \\ -H "Content-Type: application/json" \\ -H "MCP-Protocol-Version: 2025-06-18" \\ -H "MCP-Session-Id: your-session-id" \\ -d '{ "jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": { "name": "linear_model", "arguments": { "formula": "sales ~ marketing + season", "data": "sales,marketing,season\\n100,5,Q1\\n120,8,Q2\\n115,6,Q1\\n140,10,Q2\\n135,9,Q1\\n150,12,Q2", "format": "csv" } } }' **Result Interpretation (Both Methods):** - Every $1 in marketing spend generates $4.70 in additional sales - Seasonal effects show Q2 outperforms Q1 by $8.5 on average - Model explains 97.9% of sales variation (R² = 0.979) Time Series Forecasting Example ------------------------------- Forecast quarterly sales data. Python Package ~~~~~~~~~~~~~~ .. code-block:: python # Monthly sales data sales_data = [100, 105, 110, 108, 115, 120, 125, 130, 128, 135] # ARIMA forecasting forecast = rmcp.time_series_arima( data=sales_data, variable_name="monthly_sales", forecast_periods=3 ) print(forecast) HTTP Server ~~~~~~~~~~~ .. code-block:: bash curl -X POST https://rmcp-server-394229601724.us-central1.run.app/mcp \\ -H "Content-Type: application/json" \\ -H "MCP-Protocol-Version: 2025-06-18" \\ -H "MCP-Session-Id: your-session-id" \\ -d '{ "jsonrpc": "2.0", "id": 4, "method": "tools/call", "params": { "name": "time_series_arima", "arguments": { "data": [100, 105, 110, 108, 115, 120, 125, 130, 128, 135], "variable_name": "monthly_sales", "forecast_periods": 3 } } }' **Result Interpretation (Both Methods):** - ARIMA(1,1,1) model selected automatically - Next 3 periods forecasted: 138, 142, 145 - 95% confidence intervals provided for forecasts Customer Analytics Example -------------------------- Predict customer churn using logistic regression. Python Package ~~~~~~~~~~~~~~ .. code-block:: python # Customer data customer_data = { "churn": [0, 1, 0, 1, 0, 1], "tenure": [24, 2, 36, 6, 48, 12], "monthly_charges": [50, 80, 45, 75, 55, 70], "total_charges": [1200, 160, 1620, 450, 2640, 840] } # Logistic regression churn_model = rmcp.logistic_regression( formula="churn ~ tenure + monthly_charges + total_charges", data=customer_data ) print(churn_model) HTTP Server ~~~~~~~~~~~ .. code-block:: bash curl -X POST https://rmcp-server-394229601724.us-central1.run.app/mcp \\ -H "Content-Type: application/json" \\ -H "MCP-Protocol-Version: 2025-06-18" \\ -H "MCP-Session-Id: your-session-id" \\ -d '{ "jsonrpc": "2.0", "id": 5, "method": "tools/call", "params": { "name": "logistic_regression", "arguments": { "formula": "churn ~ tenure + monthly_charges + total_charges", "data": "churn,tenure,monthly_charges,total_charges\\n0,24,50,1200\\n1,2,80,160\\n0,36,45,1620\\n1,6,75,450\\n0,48,55,2640\\n1,12,70,840", "format": "csv" } } }' **Result Interpretation (Both Methods):** - Longer tenure significantly reduces churn risk (OR = 0.887) - Higher monthly charges increase churn risk (OR = 1.030) - Model achieves 100% accuracy on sample data Interface Comparison -------------------- =================== ============================ =========================== Aspect Python Package HTTP Server =================== ============================ =========================== **Setup** ``pip install rmcp`` Use deployed server **Authentication** Claude Desktop integration Session-based with headers **Data Input** Python objects/arrays JSON/CSV in requests **Response Format** Python objects JSON-RPC responses **Error Handling** Python exceptions JSON-RPC error codes **Visualization** Direct image display Base64-encoded images **Session State** Automatic Manual session management **Concurrency** Single process Multi-session support **Deployment** Local installation Cloud/server deployment =================== ============================ =========================== When to Use Each Interface --------------------------- **Use Python Package When:** - Working with Claude Desktop - Local data analysis workflows - Python-native applications - Single-user environments - Rapid prototyping and exploration **Use HTTP Server When:** - Web application integration - Multi-user environments - Remote access requirements - Microservices architecture - Language-agnostic clients Migration Between Interfaces ---------------------------- Converting Package Code to HTTP ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. code-block:: python # Python package code result = rmcp.correlation_analysis(data=[1,2,3,4,5], variable_name="test") # Equivalent HTTP request http_request = { "jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": { "name": "correlation_analysis", "arguments": { "data": [1, 2, 3, 4, 5], "variable_name": "test" } } } Converting HTTP Code to Package ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .. code-block:: bash # HTTP request curl -X POST server/mcp -d '{"method":"tools/call","params":{"name":"t_test","arguments":{"data1":[1,2,3],"data2":[4,5,6]}}}' # Equivalent Python package code result = rmcp.t_test(data1=[1,2,3], data2=[4,5,6]) Data Format Conversions ----------------------- CSV to JSON ~~~~~~~~~~~ .. code-block:: python # CSV format (HTTP) csv_data = "sales,marketing\\n100,5\\n120,8\\n115,6" # JSON format (both interfaces) json_data = { "sales": [100, 120, 115], "marketing": [5, 8, 6] } Array to DataFrame ~~~~~~~~~~~~~~~~~~ .. code-block:: python # Array format array_data = [100, 120, 115, 140] # Convert to DataFrame format df_data = {"values": array_data} Integration Patterns -------------------- Web Application Pattern ~~~~~~~~~~~~~~~~~~~~~~~ .. code-block:: javascript // Frontend JavaScript calling HTTP server class StatisticsService { constructor() { this.baseUrl = 'https://rmcp-server-394229601724.us-central1.run.app'; this.sessionId = null; } async analyzeData(data) { if (!this.sessionId) await this.initialize(); return await this.callTool('descriptive_stats', { data: data, variable_name: 'user_data' }); } } Python Application Pattern ~~~~~~~~~~~~~~~~~~~~~~~~~~ .. code-block:: python # Backend Python calling package directly class AnalyticsEngine: def __init__(self): self.rmcp = rmcp def analyze_sales_data(self, sales_data): return self.rmcp.descriptive_stats( data=sales_data, variable_name='sales' ) Hybrid Pattern ~~~~~~~~~~~~~~ .. code-block:: python # Use package for development, HTTP for production class StatisticsClient: def __init__(self, use_http=False): if use_http: self.client = HTTPRMCPClient() else: self.client = PackageRMCPClient() def analyze(self, data): return self.client.descriptive_stats(data) Performance Considerations -------------------------- **Python Package:** - Faster for single operations (no network overhead) - Memory efficient for large datasets - Direct R integration **HTTP Server:** - Better for concurrent users - Scalable with load balancing - Network latency for small operations **Recommendations:** - Use package for exploratory analysis - Use HTTP server for production applications - Consider hybrid approach for development → production workflow 🔗 **Related Documentation:** - **Package Setup**: :doc:`../package/user_guide/installation` - **HTTP Server Setup**: :doc:`../http-server/getting-started` - **Complete Tool Reference**: :doc:`../package/api/modules`