← ClaudeAtlas

forecasting-ch25-reference-classeslisted

Use when project duration or cost needs an outside-view reference class and a risk-based commitment, or studying forecasting book chapter 25.
KarpelesPublishing/the-art-and-science-of-forecasting · ★ 0 · Code & Development · score 72
Install: claude install-skill KarpelesPublishing/the-art-and-science-of-forecasting
# Chapter 25: Forecasting Your Own Life ## Scope and intake Use when project duration or cost needs an outside-view reference class and a risk-based commitment. Do not use for excluding unfinished or abandoned cases merely to make the reference class convenient. Ask only for unresolved material inputs: What outcome and decision percentile matter? Which projects were comparable before their outcomes were known? Are unfinished cases censored or failed? Do plans and actuals share units? ## Input contract and additional evidence CSV case_id,planned,actual,completed; planned > 0, actual >= 0, completed true/false. For incomplete projects actual is elapsed follow-up at extraction, not a completed duration. Record extraction date, class-selection criteria and abandonment status separately when relevant. Minimal **format illustration**, not sufficient training data: ```csv case_id,planned,actual,completed A,12,15,true B,12,18,false ``` ## Executable interface Exact CLI columns: `case_id,planned,actual,completed`. Supported method controls: `as_of, frequency, horizon, season, seed`. Kaplan–Meier estimates the ratio survival curve under independent right-censoring. P50/P80/P90 are the first steps crossing their cumulative probabilities, not interpolated empirical quantiles. Unsupported upper quantiles remain null. Abandonment needs a separate outcome interpretation. ## Applied procedure 1. Define the reference class by size, complexity, technology and operating context with