The relationship between increasing sample area (as a percentage of global area: x-axis, denoted by a) and number of events in a composite (z-axis, denoted by n) on noise levels indicated by the 97.5 percentile mean cloud fraction anomalies (%, y-axis) within composites. These values constitute the upper p = 0.05 confidence interval. Each data point is calculated from 10,000 Monte Carlo (MC) simulations from cloud cover anomaly data (21-day running mean subtracted) using PDFs. All axes shown are logarithmic.