Lecture 11: Between subjects vs. within subjects designs and interpreting confidence intervals Flashcards
Within subjects design
same subjects in both conditions
Why use a within-subjects design?
- Fewer subjects = lower cost
- Ensures groups are equivalent since they’re the same
- within-subject designs usually give us a smaller s = narrower CI
Drawbacks of within-subjects designs
- Order effects: practice, fatigue
- some experiments just don’t work within subjects eg. deception/surprise might not work twice, and can’t treat and not treat illness in the same person
Conclusions about the population depend on?
The confidence interval!
What would you conclude for a narrow confidence interval that includes zero?
We’re (95%) confident that the true value of population mean difference is either 0 or close to zero
What would you conclude for a wide confidence interval that includes zero?
Population mean difference could be big and positive, big and negative, small and positive, small and negative, or zero. BASICALLY ANYTHING
What would you conclude for a narrow confidence interval that doesn’t include zero by far?
We’re (95%) confident that the true value of the population mean difference is far above/below zero because all the values in the CI are far above zero
What would you conclude for a narrow confidence interval that doesn’t include zero but is near zero?
We’re confident that the true value of the population mean is either above/below zero
What would you conclude for a wide confidence interval that doesn’t include zero by far?
We’re confident that it’s just not zero, it could be a range of anything from a big and positive to a big and negative true value
What would you conclude for a wide confidence interval that doesn’t include zero but is near zero?
we’re not super confidence that the true value isn’t zero
