Visualizing Error Types with Sampling Distributions
What you'll learn: How to see Type I and Type II errors as regions in overlapping sampling distributions.
The Big Picture
Imagine two bell curves slightly overlapping on a graph. The first represents what you'd expect if the null hypothesis (H₀) is true—call this the "null distribution." The second shows what you'd see if the alternative hypothesis (H₁) is true—the "alternative distribution." Where these curves overlap is where confusion and errors happen.
Reading the Visualization
When you set a significance level (α), you draw a vertical line on the null distribution creating a rejection region. Everything beyond this line (in the tail) means "reject H₀."
Type I Error (α) appears as the shaded area under the null curve that falls into the rejection region. This represents falsely rejecting a true null hypothesis—a false positive.
Type II Error (β) appears as the shaded area under the alternative curve that falls outside the rejection region (where you fail to reject H₀). This represents failing to detect a real effect when it exists—a false negative.
Why This Matters
The visualization reveals the fundamental trade-off: Moving your rejection boundary left (stricter α) shrinks the Type I error region but grows the Type II error region. The further apart your distributions are (larger effect size), the less they overlap, and the smaller both error rates become.
This is why power analysis matters: you want distributions separated enough that both error zones stay small.
Key Takeaway: Overlapping sampling distributions turn abstract error probabilities into visible regions—Type I error lives in the null distribution's tail, Type II error lives in the alternative distribution's body where you fail to reject.