Dr. QuantMaster MCP Server

Instructions

Implementation Reference

• src/tools/index.ts:1483-1542 (handler)

  The main handler function for the 'generate_r_code' tool. It generates R code templates for various analysis types (OLS, panel FE, DID, etc.) based on the input 'analysis_type'.

  function handleGenerateRCode(args: Record<string, unknown>) {
    const analysisType = args.analysis_type as string;
    const templates: Record<string, string> = {
      ols: `
  # OLS Regression
  library(tidyverse)
  library(fixest)
  library(modelsummary)
  
  # Estimate
  model <- lm(y ~ x1 + x2 + x3, data = df)
  
  # Robust SE
  model_robust <- feols(y ~ x1 + x2 + x3, data = df, vcov = "hetero")
  
  # Diagnostics
  car::vif(model)  # Multicollinearity
  lmtest::bptest(model)  # Heteroscedasticity
  
  # Results table
  modelsummary(list("OLS" = model, "Robust" = model_robust))
  `,
      panel_fe: `
  # Panel Fixed Effects
  library(fixest)
  library(modelsummary)
  
  # Entity FE
  model_fe <- feols(y ~ x1 + x2 | id, data = panel_df, vcov = ~id)
  
  # Entity + Time FE
  model_twfe <- feols(y ~ x1 + x2 | id + year, data = panel_df, vcov = ~id)
  
  # Results
  modelsummary(list("Entity FE" = model_fe, "TWFE" = model_twfe))
  `,
      did: `
  # Difference-in-Differences
  library(fixest)
  library(did)
  
  # Basic DID
  did_model <- feols(y ~ treat:post | id + time, data = df, vcov = ~id)
  
  # Event Study
  es_model <- feols(y ~ i(time, treat, ref = -1) | id + time, data = df, vcov = ~id)
  iplot(es_model)
  
  # Callaway-Sant'Anna (staggered)
  cs_did <- att_gt(yname = "y", tname = "time", idname = "id",
                    gname = "first_treat", data = df)
  aggte(cs_did, type = "dynamic") |> ggdid()
  `
    };
  
    return {
      analysis_type: analysisType,
      r_code: templates[analysisType] || "# Analysis template not found\n# Use search_stats_knowledge for guidance"
    };
  }

• src/tools/index.ts:350-361 (registration)

  The tool registration in the exported tools array, including name, description, and input schema for validation.

    name: "generate_r_code",
    description: "R 코드 생성 (분석, 진단, 시각화)",
    inputSchema: {
      type: "object",
      properties: {
        analysis_type: { type: "string", description: "분석 유형" },
        variables: { type: "object", description: "변수 정보" },
        options: { type: "object", description: "추가 옵션" },
      },
      required: ["analysis_type"],
    },
  },

• src/tools/index.ts:832-833 (handler)

  The switch case in handleToolCall that routes calls to 'generate_r_code' to the specific handler function.

  case "generate_r_code":
    return handleGenerateRCode(args);