R Econometrics MCP Server

""" RMCP Streamlit Cloud App Interactive econometric analysis interface powered by Claude AI. This is a cloud-compatible version that demonstrates RMCP capabilities without requiring the local R/RMCP installation. """ import json import os import tempfile from datetime import datetime, timedelta import numpy as np import pandas as pd import streamlit as st # Anthropic Claude API try: import anthropic ANTHROPIC_AVAILABLE = True except ImportError: ANTHROPIC_AVAILABLE = False st.error("Please install the Anthropic SDK: `pip install anthropic`") # Page config st.set_page_config( page_title="RMCP - R Econometrics with Claude AI", page_icon="📊", layout="wide" ) # Header st.title("📊 RMCP - R Econometrics with Claude AI") st.markdown("*Advanced econometric analysis with R, powered by Claude AI assistance*") # Information about this demo st.info( """ 🌟 **Welcome to the RMCP Demo!** This is a demonstration version running on Streamlit Community Cloud. For the full experience with R-powered econometric analysis, you can: - 📦 Install RMCP locally: `pip install rmcp` - 🔧 Run locally: `rmcp start` - 🖥️ Use Claude Desktop with MCP integration **Current demo features:** - 🤖 Claude AI assistance for econometric questions - 📊 Sample data generation and exploration - 📈 Basic statistical analysis with Python - 💡 Econometric method recommendations """ ) # Sidebar for API key st.sidebar.title("🔧 Configuration") # Claude API Key input if ANTHROPIC_AVAILABLE: api_key = st.sidebar.text_input( "Claude API Key", type="password", help="Enter your Anthropic Claude API key from https://console.anthropic.com/", ) if api_key: client = anthropic.Anthropic(api_key=api_key) st.sidebar.success("✅ Claude API connected") else: st.sidebar.warning("⚠️ Please enter your Claude API key") client = None else: client = None # RMCP Tools Overview st.sidebar.subheader("🛠️ RMCP Tool Categories") st.sidebar.markdown( """ **📈 Descriptive Statistics** - Summary statistics, correlations, outliers **📊 Regression Analysis** - Linear, logistic, panel models **🧪 Statistical Tests** - T-tests, ANOVA, normality tests **📉 Time Series & Econometrics** - ARIMA, VAR, panel regression, IV **🔄 Data Transformations** - Lags, differences, winsorization **📊 Visualizations** - Plots, heatmaps, diagnostics **🤖 Machine Learning** - Clustering, trees, random forest """ ) # Sample data generation st.subheader("📚 Sample Economic Data") col1, col2, col3 = st.columns(3) with col1: if st.button("📊 Generate Economic Panel Data"): # Create sample economic dataset np.random.seed(42) countries = ["USA", "GER", "JPN", "UK", "FRA", "ITA", "CAN", "AUS"] years = list(range(2000, 2024)) data_rows = [] for country in countries: for year in years: # Create realistic economic relationships base_gdp = np.random.normal(2.5, 1.0) unemployment = max(0.5, np.random.normal(5.5, 2.0)) inflation = max(0, np.random.normal(2.5, 1.5)) # GDP growth negatively correlated with unemployment gdp_growth = ( base_gdp - 0.3 * (unemployment - 5.5) + np.random.normal(0, 0.5) ) data_rows.append( { "country": country, "year": year, "gdp_growth": round(gdp_growth, 2), "unemployment": round(unemployment, 2), "inflation": round(inflation, 2), "interest_rate": round(max(0, np.random.normal(2.0, 1.0)), 2), "debt_to_gdp": round(max(20, np.random.normal(60, 20)), 1), } ) sample_data = pd.DataFrame(data_rows) st.session_state["sample_data"] = sample_data st.success("✅ Economic panel data generated!") st.dataframe(sample_data.head(10)) # Basic analysis st.write("**Quick Analysis:**") st.write(f"- {len(sample_data)} observations across {len(countries)} countries") st.write(f"- Time period: {min(years)}-{max(years)}") st.write(f"- Average GDP growth: {sample_data['gdp_growth'].mean():.2f}%") st.write(f"- Average unemployment: {sample_data['unemployment'].mean():.2f}%") with col2: if st.button("📈 Generate Time Series Data"): # Create sample time series np.random.seed(123) dates = pd.date_range(start="2020-01-01", end="2023-12-31", freq="D") # Simulated economic indicators with trends and seasonality trend = np.linspace(100, 120, len(dates)) seasonal = 5 * np.sin(2 * np.pi * np.arange(len(dates)) / 365.25) noise = np.random.normal(0, 2, len(dates)) ts_data = pd.DataFrame( { "date": dates, "stock_index": trend + seasonal + noise, "gdp_index": trend * 0.8 + np.random.normal(0, 1, len(dates)), "unemployment_rate": 5 + np.sin(2 * np.pi * np.arange(len(dates)) / 365.25) + np.random.normal(0, 0.5, len(dates)), "inflation_rate": 2.5 + 0.3 * np.sin(2 * np.pi * np.arange(len(dates)) / 365.25) + np.random.normal(0, 0.3, len(dates)), } ) st.session_state["ts_data"] = ts_data st.success("✅ Time series data generated!") st.line_chart(ts_data.set_index("date")[["stock_index", "gdp_index"]]) with col3: if st.button("🏦 Generate Financial Data"): # Create sample financial dataset np.random.seed(456) n = 1000 # Simulated bank data financial_data = pd.DataFrame( { "bank_id": range(1, n + 1), "total_assets": np.random.lognormal(15, 1.5, n), "total_loans": np.random.lognormal(14, 1.2, n), "deposits": np.random.lognormal(14.5, 1.3, n), "net_income": np.random.normal(50, 25, n), "loan_loss_provision": np.random.exponential(5, n), "tier1_capital_ratio": np.random.normal(12, 3, n), "roa": np.random.normal(1.2, 0.8, n), "roe": np.random.normal(8, 4, n), "region": np.random.choice(["North", "South", "East", "West"], n), } ) # Ensure realistic relationships financial_data["total_loans"] = np.minimum( financial_data["total_loans"], financial_data["total_assets"] * 0.8 ) financial_data["tier1_capital_ratio"] = np.maximum( 8, financial_data["tier1_capital_ratio"] ) st.session_state["financial_data"] = financial_data st.success("✅ Financial dataset generated!") st.dataframe(financial_data.head(10)) # Data upload section st.subheader("📁 Data Upload & Analysis") uploaded_file = st.file_uploader( "Upload your dataset (CSV format)", type=["csv"], help="Upload a CSV file for analysis", ) # Use uploaded data or sample data data = None if uploaded_file is not None: try: data = pd.read_csv(uploaded_file) st.success(f"✅ Data loaded: {data.shape[0]} rows, {data.shape[1]} columns") except Exception as e: st.error(f"❌ Error loading data: {e}") elif "sample_data" in st.session_state: if st.checkbox("Use generated economic panel data"): data = st.session_state["sample_data"] st.success( f"✅ Using sample data: {data.shape[0]} rows, {data.shape[1]} columns" ) elif "ts_data" in st.session_state: if st.checkbox("Use generated time series data"): data = st.session_state["ts_data"] st.success( f"✅ Using time series data: {data.shape[0]} rows, {data.shape[1]} columns" ) elif "financial_data" in st.session_state: if st.checkbox("Use generated financial data"): data = st.session_state["financial_data"] st.success( f"✅ Using financial data: {data.shape[0]} rows, {data.shape[1]} columns" ) # Data exploration if data is not None: with st.expander("📋 Data Preview & Info"): col1, col2 = st.columns(2) with col1: st.write("**📊 Data Preview:**") st.dataframe(data.head()) with col2: st.write("**ℹ️ Data Info:**") st.write(f"Shape: {data.shape}") st.write("**Column Types:**") st.write(data.dtypes) if data.isnull().sum().sum() > 0: st.write("**Missing Values:**") st.write(data.isnull().sum()) # Basic statistical analysis numeric_cols = data.select_dtypes(include=[np.number]).columns.tolist() if numeric_cols: st.subheader("📈 Basic Statistical Analysis") selected_vars = st.multiselect( "Select variables for analysis:", numeric_cols, default=numeric_cols[:4] if len(numeric_cols) >= 4 else numeric_cols, ) if selected_vars: # Summary statistics st.write("**📊 Summary Statistics:**") summary_stats = data[selected_vars].describe() st.dataframe(summary_stats) # Correlation matrix if len(selected_vars) > 1: st.write("**🔗 Correlation Matrix:**") corr_matrix = data[selected_vars].corr() st.dataframe(corr_matrix) # Simple visualization st.write("**📈 Data Visualization:**") chart_type = st.selectbox( "Chart type:", ["Line Chart", "Scatter Plot", "Histogram"] ) if chart_type == "Line Chart" and len(selected_vars) >= 2: st.line_chart(data[selected_vars]) elif chart_type == "Scatter Plot" and len(selected_vars) >= 2: x_var = st.selectbox("X variable:", selected_vars) y_var = st.selectbox( "Y variable:", [v for v in selected_vars if v != x_var] ) if x_var and y_var: st.scatter_chart(data, x=x_var, y=y_var) elif chart_type == "Histogram": hist_var = st.selectbox("Variable for histogram:", selected_vars) if hist_var: st.bar_chart(data[hist_var].value_counts().head(20)) # Claude AI Assistant if client: st.subheader("🤖 Claude AI Econometric Assistant") # Predefined econometric questions st.write("**💡 Quick Questions:**") quick_questions = [ "What econometric method should I use for panel data analysis?", "How do I test for multicollinearity in my regression?", "What's the difference between fixed effects and random effects?", "How do I interpret instrumental variables regression results?", "What diagnostic tests should I run after regression analysis?", "How do I handle heteroskedasticity in my model?", "What's the best way to test for stationarity in time series?", "How do I choose the right ARIMA model order?", ] selected_question = st.selectbox("Choose a question:", [""] + quick_questions) user_question = st.text_area( "Or ask your own question about econometrics:", value=selected_question, placeholder="e.g., 'How do I handle endogeneity in my regression model?'", ) if st.button("💬 Ask Claude") and user_question: try: with st.spinner("🤔 Claude is thinking..."): # Prepare context about the data if available data_context = "" if data is not None: numeric_cols = data.select_dtypes( include=[np.number] ).columns.tolist() categorical_cols = data.select_dtypes( include=["object"] ).columns.tolist() data_context = f""" User's Dataset Context: - Shape: {data.shape[0]} rows, {data.shape[1]} columns - Numeric columns: {', '.join(numeric_cols)} - Categorical columns: {', '.join(categorical_cols)} - Sample data preview: {data.head(3).to_dict()} """ econometric_context = """ You are an expert econometrician helping with RMCP (R Model Context Protocol) - a tool that provides advanced econometric analysis capabilities including: Key RMCP capabilities: - Linear and nonlinear regression models - Panel data analysis (fixed effects, random effects, dynamic panels) - Time series analysis (ARIMA, VAR, cointegration tests) - Instrumental variables and causal inference methods - Advanced diagnostics (heteroskedasticity, autocorrelation, specification tests) - Machine learning approaches for econometrics - Comprehensive visualization tools Please provide practical, actionable advice for econometric analysis. """ message = client.messages.create( model="claude-3-5-sonnet-20241022", max_tokens=1500, messages=[ { "role": "user", "content": f"{econometric_context}\n\n{data_context}\n\nUser question: {user_question}", } ], ) st.write("**Claude's Response:**") st.write(message.content[0].text) except Exception as e: st.error(f"❌ Error communicating with Claude: {e}") else: st.info("🔑 Enter your Claude API key in the sidebar to enable AI assistance") # RMCP Installation Instructions st.subheader("🚀 Get the Full RMCP Experience") st.markdown( """ To access all RMCP's powerful econometric capabilities locally: ### 📦 Installation ```bash pip install rmcp ``` ### 🔧 Usage ```bash # Start RMCP server rmcp start # List available tools rmcp list-capabilities # Use with Claude Desktop (MCP integration) # Add to claude_desktop_config.json: { "mcpServers": { "rmcp": { "command": "rmcp", "args": ["start"] } } } ``` ### 🛠️ Available Analysis Tools - **33+ statistical and econometric tools** - **R-powered backend** for advanced analysis - **Professional visualizations** with ggplot2 - **Comprehensive diagnostics** and model validation - **Export capabilities** for research papers ### 📖 Documentation - GitHub: [github.com/your-repo/rmcp](https://github.com) - Documentation: Full API reference and examples - Examples: Real-world econometric analysis workflows """ ) # Footer st.markdown("---") st.markdown( """ **About RMCP:** RMCP (R Model Context Protocol) bridges the gap between AI assistants and professional econometric analysis. Built for researchers, analysts, and students who need robust statistical computing with modern AI assistance. *This demo showcases RMCP's potential. Install locally for the complete R-powered experience.* """ ) # Download links for sample data if "sample_data" in st.session_state: st.download_button( "📥 Download Economic Panel Data", data=st.session_state["sample_data"].to_csv(index=False), file_name="economic_panel_data.csv", mime="text/csv", ) if "ts_data" in st.session_state: st.download_button( "📥 Download Time Series Data", data=st.session_state["ts_data"].to_csv(index=False), file_name="time_series_data.csv", mime="text/csv", ) if "financial_data" in st.session_state: st.download_button( "📥 Download Financial Data", data=st.session_state["financial_data"].to_csv(index=False), file_name="financial_data.csv", mime="text/csv", )