Testing
Lecture 21
Dr. Elijah Meyer + Konnie Huang
Duke University
STA 199 - Fall 2022
November 9th, 2022
Announcements
– HW 5 (Due Thursday)
– Small typo on HW 5 (see Slack)
– You may see some hints posted as well….
– Drop 1 Peer Feedback
Goals
– Compare two proportions, and do a hypothesis test
– Compare two means, and do a bootstrap interval
– Prediction and bootstrap interval for the slope
Warm Up: Can we always make inference?
– What assumption do we need to make about our data in order to trust results from our bootstrap methods?
Assumption - Independence
– What is independence?
- One observation does not effect or influence another
Independence Example
– 50 Airbnb listings in 2020 for Asheville, NC.
![]()
How do we ensure Independnece
– Random Sample - Every observation has an equal probability of being selected
– A random sample ensures that our sample is representative of the entire population
![]()
If we don’t have a random sample …
– Our results are hard to trust
– Good models and procedures can not fix bad data
For your project
– Today, we go through how to bootstrap for a slope coefficient
– Can use other bootstrap methods for other questions you may be interested in
– Do not use if data are not representitative
In Summary
