Artifacts that teach this idea, ranked by trust: tier, then passing built-in checks and accessibility, then likes.
Students run a simulated tutoring experiment again and again, and see that any single RCT estimate misses the true effect while the average of many lands on it. They also see the spread of estimates shrink with the square root of n. Interactive dials drive R code cells, and students then complete the key calculations themselves.
Alexis Diamond · ♥ 0