forecasting-ch07-monte-carlolisted

# Chapter 7: The Casino at Los Alamos ## Scope and intake Use when uncertain inputs must be propagated through a calculation or simulation precision needs auditing. Do not use for creating probabilities by assigning arbitrary distributions to unsupported low/base/high scenarios. Ask only for unresolved material inputs: What outcome and units matter? Which input distributions are measured or elicited? Which dependencies matter? What precision is needed for the decision? ## Input contract and additional evidence CSV component,mean,sd for positive lognormal components, with mean > 0 and sd >= 0 in common additive units. Config samples,seed and correlation assumption. Arithmetic-scale mean/sd are not log-scale normal parameters. Other distributions require an explicitly adapted model. Minimal **format illustration**, not sufficient training data: ```csv component,mean,sd Labor,100,20 Materials,50,15 ``` ## Executable interface Exact CLI columns: `component,mean,sd`. Supported method controls: `as_of, correlation, frequency, horizon, samples, season, seed`. Positive arithmetic means and nonnegative SDs define lognormal components. correlation is shared latent-normal correlation in [0,1), not the resulting components’ Pearson correlation. The adapter simulates totals directly; it does not run MCMC or report budget exceedance unless added separately. ## Applied procedure 1. Define the simulation function and input provenance. Specify marginal distributions and dependence