Every time we make a decision based on a Statistic Hypothesis, there are four possible outcomes, for depending on our decision and the Null Hypothesis:

Decision / Reality𝐻0 is True𝐻0 is False
Reject 𝐻0Type I Error (false positive)Correct decision (power)
Do not reject 𝐻0Correct decision (confidence)Type II Error (false negative)

Definitions:

Type I:

Rejecting 𝐻0 when it is actually true. Error 𝛼

𝛼=𝑃(Reject 𝐻0𝐻0 is true)significance level.

Type II Error

Not rejecting 𝐻0 when it is actually false. Error 𝛽

𝛽=𝑃(Do not reject 𝐻0𝐻0 is false).

Important Concepts

  • Power of the test: 1𝛽, the probability of correctly rejecting a false 𝐻0.
  • Confidence level: 1𝛼, the probability of correctly not rejecting a true 𝐻0.

Important relationships:

  • 𝛼 and 𝛽 do not sum to 1. They are inversely related (generally, reducing one increases the other).
  • The only way to reduce both simultaneously is to increase the sample size 𝑛.
  • The choice of 𝛼 depends on the context: which error is more costly?

Example from the document (parachutist):

  • 𝐻0: The equipment is correctly installed (I jump).
  • 𝐻𝑎: The equipment is faulty (I do not jump).
  • Type I error: I think the equipment is faulty, but it is actually fine → I lose the chance to jump (less severe).
  • Type II error: I think the equipment is fine, but it is faulty → I jump and die (more severe). Here the Type II error is more serious, so we might tolerate a higher 𝛼 to keep 𝛽 low.