Presentation Transcript
Hypothesis Testing & Construction: Hypothesis Testing & Construction Chee-Wee Tan P1112
What are hypotheses?: What are hypotheses? Hypothesis testing – put forward a statement / question for testing whether its true.
H1 – Experimental Hypothesis
H0 – Null hypothesis
What are hypotheses?: What are hypotheses? H1 – Watching day-time TV increases a person’s boredom
H0 – Watching day-time TV does not increase a person’s boredom.
Why use the null hypothesis?: Why use the null hypothesis? You can never prove a hypothesis but only disprove it.
A scientific theory is falsifiable. Spot the Vulcan contest
How to test a hypothesis?: How to test a hypothesis? If probability of results happening by chance decreases, more confident of rejecting null hypothesis.
Convention: if result happening by chance is less than 5%, reject null hypothesis.
How to test a hypothesis? : How to test a hypothesis? Steps:
Construct null hypothesis and experimental hypothesis.
Administer intervention to one of the groups with the other acting as control.
Calculate the probability of an effect happening by chance.
If less than 5%, reject null hypothesis.
How to test a hypothesis: How to test a hypothesis H0: Students with enforced laziness does not have worse exam grades than students without enforced laziness. Class Enforced
Laziness Control Exams P =0.3 Compare
Determining if samples are different: Determining if samples are different Depends on:
Research design
Statistical test
2 types of variation
Systematic
Unsystematic
Test statistic and variations: Test statistic and variations Test statistic: number with known characteristic. Examples, t and F.
Tests statistic = Systematic variance
Unsystematic variance
The larger the test statistic, the better.
Two mistakes we must know: Two mistakes we must know Effect No effect Effect No effect Findings Actual
Type I & II errors: Type I & II errors Type I error (α-level)
Usually set at 0.05
Type II error (β-level)
Maximum acceptable level = 0.2
Increasing α will decrease β, vice versa Where is my sweater?
Effect size: Effect size If an effect exists, doesn’t mean it’s important or meaningful.
Need to find effect size –
The magnitude of the effect
Used in meta-analysis
Statistics: d, r, q, g, h, w, f & f2
Summary: Summary Hypothesis testing
Null hypothesis
How hypothesis testing works
Test statistics & variations
Type I & II errors
Effect size Resistance is futile, assimilate knowledge