Hypothesis Testing Flashcards
(15 cards)
What is a null hypothesis (𝐻0)
The null hypothesis is a formal statement that there is no difference, no relationship, or no effect in the population
It’s what you’re testing against
You assume it’s true until there’s enough evidence to reject it
What is an Alternative Hypothesis (𝐻1 or 𝐻𝑎)
The alternative hypothesis is what you are trying to support with evidence. It claims there is a real effect or difference.
It opposes the null hypothesis.
If the data supports it strongly enough, you reject 𝐻0 in favour of 𝐻1
What is a 1-sided test
Tests for a specific direction (greater than or less than)
Use when theory or question expects change in one direction
What is a 2-sided test
Tests for any difference, in either direction
More common in science unless a strong reason exists
What is a critical value
the cutoff point beyond which you reject the null hypothesis based on your significance level (α)
It depends on your test type (1- or 2-sided) and confidence level
For α = 0.05 (95% CI), the critical z-value is:
±1.96 for 2-sided test
1.645 for one-sided (either left or right)
What is a p-value and what does it tell you
The p-value is the probability of observing your data if the null hypothesis is true
Low p-value (< α): Evidence against 𝐻0
High p-value: Not enough evidence to reject 𝐻0
What is Type I error
happens when you reject a true null hypothesis — a false positive
You claim there is an effect, when there really isn’t
The significance level 𝛼 is the probability of making a Type I error
What does a standard 𝛼 of 0.05 tell you about the risk of a Type I error
You’re okay with a 5% risk of a false positive
What is a Type II error and how does it relate to statistical power
A Type II Error occurs when you fail to reject a false null hypothesis — a false negative.
You miss a real effect.
Denoted by β.
The probability of detecting a true effect is called statistical power.
How do you calculate statistical power
Power = 1 − β
What does a high statistical power mean in term of a Type II error
Higher power = lower chance of Type II error = better test sensitivity
Typical target for power: 80%
What does ‘correlation is not causation’ mean
Just because two variables are related (correlated) doesn’t mean one causes the other.
There could be
1. Coincidence
2. A third variable (confounder)
3. A reverse relationship
What is a confounding variable
A variable that influences both the independent and dependent variables, making it seem like they are related when the real cause is the confounding variable.
What is Simpson’s paradox
A situation where a trend appears in individual groups, but reverses when the data is combined.
Shows how aggregated data can be misleading
What is Occam’s razor
A philosophical principle used in science and modelling: “The simplest explanation that fits the data is usually the best one.”
Avoid overfitting or adding unnecessary complexity.
Prefer parsimony - fewer assumptions or variables.
