LECTURE HUB
Lecture 6 - Confidence intervals
🚦 Status🟠 Studying
🧠 Mastery🐥 Apprentice
📅 Last Review8 Jun 2025
🎯KEY TAKEAWAYS
- An interval is a range of plausible values for the true number, not the spread of the data.
- The 95 % belongs to the procedure: do this every week and about 19 intervals in 20 will contain the truth. This one either does or does not.
- Halving the width costs four times the sample.
- I should not have missed last week.
⚡QUICK JOTS
- 18:35 everybody else has last week's handout. I have a launch.
- standard error used as if I knew it. Problem set 2 did cover it. Sort of.
- interval = estimate ± (how many SEs) × SE
- 1.96?? from the normal table. ask Rasmus for the 11 March notes
- 95 % describes the method, not this one interval (Hale said it three times, slowly)
- wider sample → narrower interval, but by a square root. precision is expensive
- 20:10 lost the thread at the t-distribution. wrote down the word and nothing else
☑️HOMEWORK
- Borrow Rasmus's notes from 11 March and copy the part on sampling distributions
- Rewrite 🔤 Confidence interval with the sentence I used to say, crossed out
- Work one interval by hand with 📐 Confidence interval for a proportion
- Find out when the t-distribution matters and when I can ignore it