Errors and Consequences

Statistics · Axiom Academy

Why perfect decisions are impossible and how we navigate uncertainty. Imagine you're designing a medical test for a disease. Move the slider to adjust how "strict" your test is. Different situations demand different priorities. Click on each scenario to see which error is more costly. Understanding the Four Outcomes Every binary decision leads to one of four possible outcomes. This is called a confusion matrix. Type I Error (False Positive): Rejecting something true. Seeing a pattern that isn't there. "False alarm." Type II Error (False Negative): Accepting something false. Missing a real pattern. "Missed detection." Adjust the parameters below to see how they affect your ability to detect real effects. When distributions overlap (uncertainty exists), no decision rule can perfectly separate groups. You must choose which errors you can tolerate. Being more conservative (reducing false positives) makes you miss more real effects (increasing false negatives). The optimal balance depends on the costs and context. To maintain low error rates while detecting subtle effects, you need larger samples, better measurements, or stronger effects. There's no free lunch. Every statistical test involves choosing α (acceptable false positive rate) and ensuring adequate power. These choices reflect your priorities about which mistakes are more costly.

This is the written version of the interactive lesson above. See the full Statistics course.