The 4 basic elements of a hypothesis test
Clarify your statistical thinking in applied statistics
Hypothesis testing is everywhere in applied statistics, data science, and business intelligence - but it is widely misused and misunderstood. I am sympathetic to why it confuses people; a hypothesis test can have a lot of detail and complexity, and the logic behind it is not intuitive.
However, every hypothesis test has 4 basic elements, and I always return to them to clarify a convoluted test.
What is the null hypothesis, H₀?
What is the alternative hypothesis, Hₐ?
What is your test statistic, T? By choosing your test statistic, you also choose the probability distribution of your test statistic.
What is your rejection rule? This usually involves defining the significance level, α. A typical rejection rule is to reject the null hypothesis when the P-value is smaller than the significance level.
If you cannot state these 4 elements precisely, then you don’t fully understand your hypothesis test. Make sure that you can articulate them with clarity.
In a future article, I will use an example to illustrate these 4 elements. Please stay tuned!