Incident Severity Power-Law Analyzer

Real incident-severity datasets — earthquake energy, wildfire acreage, power-outage duration, injury counts — overwhelmingly follow a power-law or log-normal distribution: enormous numbers of minor events and a thin, naturally decaying tail of severe ones. Plotted on log-log axes this shows up as an almost straight line. This simulator synthesizes a batch of incident reports from a true power-law generator, lets you dial in an "embellishment bias" that inflates a fraction of reports the way exaggerated retellings do, and fits both a maximum-likelihood exponent and a log-log linear regression to the result — so you can see, numerically, when a severity distribution stops looking natural and starts looking padded.