analysis_bundle.json

{
  "artifacts": [
    {
      "alt_text": "聚合效应图：treatment - control 方向的 ANCOVA 校正差与 Welch 未校正差，均带 95% 置信区间和零效应参考线，不包含行级数据点。",
      "byte_size": 47882,
      "chart_id": "CH-F675F0E546C6",
      "evidence_ids": [
        "E-7C87BB6C88EB",
        "E-B93CD9DC7751"
      ],
      "file_name": "effect_estimates.png",
      "height_px": 692,
      "mime_type": "image/png",
      "plot_spec_sha256": "bcb85b1a1aa4dd3da530caaef00319e13d15d2de52ec3c1fff42e1f34c0307d1",
      "schema_version": "1.0",
      "sha256": "d4e061cde8f080dd4d053d746b13448b6d1e07d786441aa51585c5cd95ee523c",
      "width_px": 1195
    }
  ],
  "created_at_utc": "2026-08-15T12:19:05.354913+00:00",
  "dataset": {
    "raw_data_embedded": false,
    "sha256": "db7ce30ae0fdc9d455edfd6f107f974215aa7fc91209f73a7d1afdf208b9062c",
    "source_columns": 10,
    "source_name": "synthetic_trial.csv",
    "source_rows": 240
  },
  "design": {
    "analysis_population": "intention_to_treat",
    "confidence_level": 0.95,
    "contrast_level": "treatment",
    "covariate_timing": {
      "baseline_sbp": "pre_treatment"
    },
    "covariates": [
      "baseline_sbp"
    ],
    "expected_cell_count": "unknown",
    "group_count": 2,
    "normality": "unknown",
    "objective": "group_difference",
    "outcome": "followup_sbp",
    "outcome_type": "continuous",
    "overdispersion": "unknown",
    "paired": false,
    "predictor": "group",
    "predictor_type": "categorical",
    "question": "在该模拟随机对照试验中，治疗组与对照组的随访收缩压是否存在差异？在考虑基线收缩压后，结论是否仍然成立？",
    "randomized": true,
    "reference_level": "control",
    "repeated_measures": false,
    "subject_id": null,
    "time_variable": null
  },
  "evidence": [
    {
      "dataset_sha256": "db7ce30ae0fdc9d455edfd6f107f974215aa7fc91209f73a7d1afdf208b9062c",
      "diagnostics": {
        "adjusted_r_squared": 0.6933977353949026,
        "automatic_normality_switch_used": false,
        "automatic_outlier_removal_used": false,
        "breusch_pagan": {
          "automatic_method_switch_used": false,
          "f_p_value": 0.39180358417871386,
          "f_statistic": 0.9412079694610557,
          "lm_p_value": 0.38821500440602497,
          "lm_statistic": 1.89239191553586
        },
        "condition_number": 27.686553261957215,
        "max_leverage": 0.06351876305511386,
        "model_rank": 3,
        "parameter_count": 3,
        "r_squared": 0.6963039179977945,
        "residual_degrees_of_freedom": 209.0,
        "residual_skewness": 0.020964670092000866,
        "slope_homogeneity": [
          {
            "confidence_interval": {
              "lower": -0.20076188411191448,
              "upper": 0.17214619533110626
            },
            "confidence_level": 0.95,
            "covariate": "baseline_sbp",
            "degrees_of_freedom": 208.0,
            "interpretation_guardrail": "Do not treat a non-significant interaction as proof of equal slopes.",
            "p_value": 0.8799005631673987,
            "slope_difference": -0.014307844390404112,
            "standard_error": 0.09457784406684967,
            "status": "estimated"
          }
        ]
      },
      "estimates": {
        "coefficients": {
          "baseline_sbp": 1.022568557455116,
          "intercept": 131.25642610197872,
          "treatment_vs_control": -5.6069303056318915
        },
        "contrast": {
          "adjusted_mean_difference": -5.6069303056318915,
          "confidence_interval": {
            "lower": -7.935143502095155,
            "upper": -3.278717109168628
          },
          "confidence_level": 0.95,
          "direction": "treatment - control",
          "standard_error_hc3": 1.1810071281941852
        },
        "contrast_group": {
          "adjusted_mean": {
            "confidence_interval": {
              "lower": 124.15459595973232,
              "upper": 127.14439563296133
            },
            "estimate": 125.64949579634683,
            "standard_error_hc3": 0.758301415720852
          },
          "level": "treatment"
        },
        "reference_group": {
          "adjusted_mean": {
            "confidence_interval": {
              "lower": 129.47996138363294,
              "upper": 133.0328908203245
            },
            "estimate": 131.25642610197872,
            "standard_error_hc3": 0.9011277397357316
          },
          "level": "control"
        }
      },
      "evidence_id": "E-7C87BB6C88EB",
      "input_spec": {
        "confidence_level": 0.95,
        "contrast_direction": "treatment - control",
        "contrast_level": "treatment",
        "covariance_estimator": "HC3",
        "covariate_center_values": {
          "baseline_sbp": 132.06603773584905
        },
        "covariate_timing": {
          "baseline_sbp": "pre_treatment"
        },
        "covariates": [
          "baseline_sbp"
        ],
        "group": "group",
        "missing_data_policy": "available_case_for_required_columns",
        "outcome": "followup_sbp",
        "outlier_policy": "no_automatic_removal",
        "reference_level": "control",
        "use_t_distribution": true
      },
      "method_code": "ancova_linear_model",
      "role": "primary",
      "sample_flow": {
        "by_group": {
          "control": {
            "excluded_rows": 18,
            "included_rows": 102,
            "source_rows": 120
          },
          "treatment": {
            "excluded_rows": 10,
            "included_rows": 110,
            "source_rows": 120
          }
        },
        "excluded_rows": 28,
        "included_rows": 212,
        "missing_by_column": {
          "baseline_sbp": 0,
          "followup_sbp": 28,
          "group": 0
        },
        "required_columns": [
          "group",
          "followup_sbp",
          "baseline_sbp"
        ],
        "source_rows": 240
      },
      "schema_version": "1.0",
      "test": {
        "alternative": "two-sided",
        "covariance_estimator": "HC3",
        "degrees_of_freedom": 209.0,
        "p_value": 3.817575932819937e-06,
        "statistic": -4.747583796725384,
        "statistic_name": "t"
      },
      "tool_name": "run_ancova",
      "tool_version": "1.0.0",
      "warnings": [
        "未自动插补缺失值，也未自动删除异常值。",
        "该方法按所需字段执行可用病例分析：240 行中纳入 212 行，排除 28 行。",
        "requested_population=intention_to_treat，但 realized_population=available_case；当前结果不得声称已完整实现 ITT。",
        "组间排除率相差 6.7%，需评估差异性失访偏倚。"
      ]
    },
    {
      "dataset_sha256": "db7ce30ae0fdc9d455edfd6f107f974215aa7fc91209f73a7d1afdf208b9062c",
      "diagnostics": {
        "automatic_normality_switch_used": false,
        "automatic_outlier_removal_used": false,
        "group_variance_ratio_max_over_min": 1.228318838550478
      },
      "estimates": {
        "contrast": {
          "cohen_d_pooled_sd": -0.45564343307939037,
          "confidence_interval": {
            "lower": -10.842458427218284,
            "upper": -2.734939077238087
          },
          "confidence_level": 0.95,
          "direction": "treatment - control",
          "hedges_g_pooled_sd": -0.4540138715154302,
          "mean_difference": -6.788698752228186,
          "standard_error": 2.055986039092505
        },
        "contrast_group": {
          "level": "treatment",
          "mean": 125.08090909090907,
          "mean_confidence_interval": {
            "confidence_level": 0.95,
            "lower": 122.4082892188539,
            "upper": 127.75352896296425
          },
          "n": 110,
          "standard_deviation": 14.142852413686903
        },
        "reference_group": {
          "level": "control",
          "mean": 131.86960784313726,
          "mean_confidence_interval": {
            "confidence_level": 0.95,
            "lower": 128.7908519185382,
            "upper": 134.9483637677363
          },
          "n": 102,
          "standard_deviation": 15.674459197427483
        }
      },
      "evidence_id": "E-B93CD9DC7751",
      "input_spec": {
        "alternative": "two-sided",
        "confidence_level": 0.95,
        "contrast_direction": "treatment - control",
        "contrast_level": "treatment",
        "group": "group",
        "missing_data_policy": "available_case_for_required_columns",
        "outcome": "followup_sbp",
        "outlier_policy": "no_automatic_removal",
        "reference_level": "control"
      },
      "method_code": "welch_t_test",
      "role": "sensitivity",
      "sample_flow": {
        "by_group": {
          "control": {
            "excluded_rows": 18,
            "included_rows": 102,
            "source_rows": 120
          },
          "treatment": {
            "excluded_rows": 10,
            "included_rows": 110,
            "source_rows": 120
          }
        },
        "excluded_rows": 28,
        "included_rows": 212,
        "missing_by_column": {
          "followup_sbp": 28,
          "group": 0
        },
        "required_columns": [
          "group",
          "followup_sbp"
        ],
        "source_rows": 240
      },
      "schema_version": "1.0",
      "test": {
        "alternative": "two-sided",
        "degrees_of_freedom": 203.5589436285064,
        "equal_variance_assumed": false,
        "p_value": 0.001134077491786515,
        "statistic": -3.301918701366601,
        "statistic_name": "t"
      },
      "tool_name": "run_welch_t_test",
      "tool_version": "1.0.0",
      "warnings": [
        "未自动插补缺失值，也未自动删除异常值。",
        "该方法按所需字段执行可用病例分析：240 行中纳入 212 行，排除 28 行。",
        "requested_population=intention_to_treat，但 realized_population=available_case；当前结果不得声称已完整实现 ITT。",
        "组间排除率相差 6.7%，需评估差异性失访偏倚。",
        "Hedges g 使用合并样本标准差进行标准化，它不是 Welch 检验统计量。"
      ]
    }
  ],
  "question": "在该模拟随机对照试验中，治疗组与对照组的随访收缩压是否存在差异？在考虑基线收缩压后，结论是否仍然成立？",
  "recommendation": {
    "assumptions_to_check": [
      "观测单位之间独立，或已在模型中正确表示相关结构。",
      "结局与连续协变量的关系在所设定的函数尺度上合理。",
      "组别与基线协变量的斜率齐性假设成立，或已显式建模交互项。",
      "模型残差均值、方差和分布假设足以支持所选推断。"
    ],
    "dataset_sha256": "db7ce30ae0fdc9d455edfd6f107f974215aa7fc91209f73a7d1afdf208b9062c",
    "primary_method": {
      "code": "ancova_linear_model",
      "label": "ANCOVA（协方差分析）",
      "purpose": "在校正预指定协变量后估计组间平均结局差异。"
    },
    "question": "在该模拟随机对照试验中，治疗组与对照组的随访收缩压是否存在差异？在考虑基线收缩压后，结论是否仍然成立？",
    "rationale": [
      "研究设计预指定了 1 个协变量：baseline_sbp。",
      "线性模型可同时给出校正组间效应、置信区间和模型诊断。",
      "研究设计已显式声明为随机分组；选择器不会仅根据列名推断这一事实。"
    ],
    "required_diagnostics": [
      "检查结局分布、异常值和各组有效样本量。",
      "检查残差图、异方差、强影响点和非线性。",
      "检查组别×基线协变量交互，并根据研究方案决定是否保留。",
      "报告各组缺失比例，比较完整与缺失个体的基线特征，并预指定缺失数据策略。",
      "normality=unknown：使用图形和模型残差诊断，不要仅根据正态性检验 p 值自动切换方法。"
    ],
    "rule_version": "2026-08-15.1",
    "sensitivity_methods": [
      {
        "code": "welch_t_test",
        "label": "Welch 独立样本 t 检验",
        "purpose": "提供两组未校正平均差。"
      }
    ],
    "status": "ready",
    "warnings": [
      "分析变量存在缺失：followup_sbp=11.7%。",
      "声明了意向性治疗分析，但结局存在缺失；仅删除缺失行不足以自动实现 ITT。"
    ]
  },
  "run_id": "RUN-C5D58104E492",
  "runtime": {
    "analysis_tool": "1.0.0",
    "chart_tool": "1.0.0",
    "matplotlib": "3.11.1",
    "numpy": "2.5.2",
    "pandas": "3.0.5",
    "platform": "Windows-11-10.0.26200-SP0",
    "python": "3.12.13",
    "scipy": "1.18.0",
    "statsmodels": "0.14.6"
  },
  "schema_version": "1.0",
  "status": "completed",
  "warnings": [
    "分析变量存在缺失：followup_sbp=11.7%。",
    "声明了意向性治疗分析，但结局存在缺失；仅删除缺失行不足以自动实现 ITT。",
    "未自动插补缺失值，也未自动删除异常值。",
    "该方法按所需字段执行可用病例分析：240 行中纳入 212 行，排除 28 行。",
    "requested_population=intention_to_treat，但 realized_population=available_case；当前结果不得声称已完整实现 ITT。",
    "组间排除率相差 6.7%，需评估差异性失访偏倚。",
    "Hedges g 使用合并样本标准差进行标准化，它不是 Welch 检验统计量。"
  ]
}