suggest_method

Instructions

Implementation Reference

• src/tools/index.ts:47-72 (registration)

  Registration of the 'suggest_method' tool in the tools array, including its name, description, and input schema.

  {
    name: "suggest_method",
    description: "연구질문과 데이터 특성에 맞는 통계 방법 추천",
    inputSchema: {
      type: "object",
      properties: {
        research_question: { type: "string", description: "연구 질문" },
        dv_type: {
          type: "string",
          enum: ["continuous", "binary", "ordinal", "count", "time_to_event", "proportion"],
          description: "종속변수 유형"
        },
        data_structure: {
          type: "string",
          enum: ["cross_sectional", "panel", "time_series", "clustered", "multilevel"],
          description: "데이터 구조"
        },
        causal_design: {
          type: "string",
          enum: ["none", "experimental", "quasi_experimental", "observational"],
          description: "인과추론 설계"
        },
      },
      required: ["research_question", "dv_type"],
    },
  },

• src/tools/index.ts:50-71 (schema)

  Input schema definition for the 'suggest_method' tool specifying parameters for research question, dependent variable type, data structure, and causal design.

  inputSchema: {
    type: "object",
    properties: {
      research_question: { type: "string", description: "연구 질문" },
      dv_type: {
        type: "string",
        enum: ["continuous", "binary", "ordinal", "count", "time_to_event", "proportion"],
        description: "종속변수 유형"
      },
      data_structure: {
        type: "string",
        enum: ["cross_sectional", "panel", "time_series", "clustered", "multilevel"],
        description: "데이터 구조"
      },
      causal_design: {
        type: "string",
        enum: ["none", "experimental", "quasi_experimental", "observational"],
        description: "인과추론 설계"
      },
    },
    required: ["research_question", "dv_type"],
  },

• src/tools/index.ts:1014-1071 (handler)

  The core handler function for 'suggest_method' that provides statistical method recommendations based on input parameters using a predefined suggestions lookup table.

  function handleSuggestMethod(args: Record<string, unknown>) {
    const dvType = args.dv_type as string;
    const dataStructure = (args.data_structure as string) || "cross_sectional";
    const causalDesign = (args.causal_design as string) || "none";
  
    const suggestions: Record<string, any> = {
      continuous: {
        cross_sectional: {
          observational: ["OLS Regression", "Robust Regression"],
          quasi_experimental: ["Propensity Score Matching", "IV/2SLS"],
          experimental: ["t-test", "ANOVA", "OLS with controls"]
        },
        panel: {
          observational: ["Panel FE/RE", "Correlated Random Effects"],
          quasi_experimental: ["DID", "Synthetic Control", "Event Study"]
        },
        time_series: {
          observational: ["ARIMA", "VAR", "VECM"],
          quasi_experimental: ["Interrupted Time Series", "Event Study"]
        }
      },
      binary: {
        cross_sectional: {
          observational: ["Logistic Regression", "Probit"],
          quasi_experimental: ["PSM + Logit", "IV Probit"]
        },
        panel: {
          observational: ["Conditional Logit (FE)", "Random Effects Logit"],
          quasi_experimental: ["DID Logit", "Linear Probability Model + FE"]
        }
      },
      count: {
        cross_sectional: {
          observational: ["Poisson", "Negative Binomial", "Zero-Inflated"]
        },
        panel: {
          observational: ["FE Poisson", "FE Negative Binomial"]
        }
      }
    };
  
    const methodList = suggestions[dvType]?.[dataStructure]?.[causalDesign] ||
      suggestions[dvType]?.[dataStructure]?.["observational"] ||
      ["검색 필요: search_stats_knowledge 사용"];
  
    return {
      research_question: args.research_question,
      dv_type: dvType,
      data_structure: dataStructure,
      causal_design: causalDesign,
      recommended_methods: methodList,
      considerations: [
        "데이터 크기와 분포 확인 필요",
        "가정 검토 (내생성, 선택편의 등)",
        "robustness check 계획 수립"
      ]
    };
  }