经济数据质量与解读审计
审查经济和市场数据的来源、定义、口径、修订与可比性,识别误读风险并形成质量评级与限制说明。
完整提示词
I want a forensic audit of the quality of economic, market and statistical data before they are used for serious conclusions. Main objective: > Determine whether the dataset or indicator is reliable, comparable and methodologically appropriate enough for the specific conclusion, and prevent wrong interpretations caused by revisions, denominators, a nominal/real mix, seasonal effects or definition changes. This is not: - automatically rejecting estimated data - only a missing-value check - a statistical analysis without understanding the methodology - correcting source data without documenting it ## 1. SOURCE ## 2. SOURCE AUTHORITY ## 3. ORIGINAL SOURCE Prefer original over copied secondary chart. ## 4. DATASET VERSION ## 5. RELEASE DATE ## 6. REFERENCE PERIOD ## 7. REVISION DATE ## 8. VINTAGE Critical for macro analysis. ## 9. DEFINITION ## 10. UNIT ## 11. CURRENCY ## 12. PRICE BASIS - current - constant - chain-linked ## 13. NOMINAL/REAL ## 14. INDEX ## 15. BASE YEAR ## 16. SEASONAL ADJUSTMENT ## 17. ANNUALIZATION ## 18. FREQUENCY ## 19. POPULATION ## 20. SAMPLE ## 21. WEIGHT ## 22. SURVEY DESIGN ## 23. RESPONSE RATE ## 24. ESTIMATION ## 25. IMPUTATION ## 26. BENCHMARK REVISION ## 27. METHODOLOGY CHANGE ## 28. SERIES BREAK ## 29. RECLASSIFICATION ## 30. GEOGRAPHY CHANGE ## 31. POPULATION CHANGE ## 32. DENOMINATOR ## 33. PER CAPITA ## 34. PPP ## 35. FX CONVERSION ## 36. INFLATION ADJUSTMENT ## 37. MISSING DATA ## 38. STRUCTURAL NA Different from missing. ## 39. ZERO Zero is not missing. ## 40. OUTLIER ## 41. DATA ERROR ## 42. TRUE EXTREME ## 43. DUPLICATE ## 44. AGGREGATION ## 45. WEIGHTED AVERAGE ## 46. SIMPLE AVERAGE ## 47. MEDIAN ## 48. RATE ## 49. LEVEL ## 50. GROWTH ## 51. CAGR ## 52. INDEX CHANGE ## 53. CONTRIBUTION ## 54. SHARE ## 55. PERCENTAGE POINT ## 56. PERCENT CHANGE Critical distinction. ## 57. STOCK/FLOW ## 58. GROSS/NET ## 59. REVISION BIAS ## 60. REAL-TIME DATA ## 61. SURVIVORSHIP ## 62. SELECTION BIAS ## 63. COVERAGE BIAS ## 64. LOOK-AHEAD BIAS ## 65. PUBLICATION LAG ## 66. COMPARABILITY Across countries/time. ## 67. METHODOLOGICAL BREAK ## 68. PROXY ## 69. PROXY VALIDITY ## 70. ALTERNATIVE DATA ## 71. SCRAPED DATA ## 72. MANUAL MAPPING ## 73. INTERPOLATION Must be explicitly disclosed. ## 74. EXTRAPOLATION ## 75. FORECAST DATA Must not be mixed with actual unnoticed. ## 76. PRELIMINARY ## 77. ESTIMATE ## 78. FINAL ## 79. SOURCE CONFLICT ## 80. RECONCILIATION ## 81. META DATA ## 82. DOCUMENTATION ## 83. LINEAGE Map: ```text original source ↓ extraction ↓ transformation ↓ calculation ↓ final metric ``` ## 84. REPRODUCIBILITY ## 85. CALCULATION CHECK ## 86. ROUNDING ## 87. SPREADSHEET ERROR ## 88. UNIT CONVERSION ## 89. CHART DISTORTION Axis/base effects. ## 90. INTERPRETATION ## 91. CAUSAL CLAIM Data quality alone cannot prove causality. ## 92. FALSE POSITIVE RULES Estimated data is not automatically bad. Revised data is not automatically unreliable. Small sample can still be useful with limits. ## 93. EVIDENCE TIERS ```text A - original source + methodology + reproducible calculation B - authoritative processed series C - reliable secondary transformation D - estimated/proxy/manual mapping E - speculative/insufficient ``` Materiality is the severity scale of this analysis: **HIGH** changes the overall conclusion or the decision; **MEDIUM** changes a key metric, driver or trend but not the conclusion; **LOW** has a limited effect and is noted for completeness. Judge it against an explicit base (the conclusion that depends on the data) and state that base. A data-quality issue is CONFIRMED only with evidence tier A or B. For the series itself: tier A supports VERIFIED; tiers B and C support USABLE_WITH_LIMITATIONS with the limitation stated; tier D stays NOT VERIFIED; tier E is UNUSABLE for the conclusion. Never present an estimate, a forecast or a scenario as a fact. ## 94. STATUS ```text VERIFIED USABLE_WITH_LIMITATIONS NOT VERIFIED INCOMPARABLE STRUCTURAL_NA UNUSABLE ``` ## 95. FINDING FORMAT ```text ID: Status: Evidence tier: Dataset/indicator: Source: Period: Definition: Issue: Affected observations: Materiality: Interpretation risk: Evidence: Can be repaired: Repair method: Remaining limitation: ``` ## 96. DATA QUALITY MATRIX | Series | Source | Definition | Comparable | Quality | |---|---|---|---|---| ## 97. SECOND PASS Check: - unit - real/nominal - percent vs pp - revisions - series breaks - missing vs zero - estimate vs actual - denominator changes - source lineage - country comparability ## 98. FINAL QUALITY GATE Confirm: - original source - version/date - definition - unit - methodology - revisions - missing data - comparability - transformations - lineage - limitations ## 99. OUTPUT `ECONOMIC_DATA_QUALITY_INTERPRETATION_AUDIT.md` # FINAL RULE Before you ask: > "What does this number mean?" first prove: ```text what the number measures how it was calculated which population it refers to in which period and whether it is comparable at all with the number you compare it to ```
分类与标签
财务理财、经济数据、数据质量、市场分析、统计解读