Data Analysis for High School Teachers

Week 6, February 15, 2000

I. Questions, follow up from last time?
II. Ozone Presentation
Yi Huang, graduate student, Department of Atmospheric Sciences, UCLA
III. Analyze the ozone data set:
i. Get acquainted with the data (15 minutes). Then
ii. Answer these questions:

1) How do each of the predictors relate to ozone levels? Can we explain this theoretically?

2) How are the variables distributed? Any interesting/problematic

features?

3) Choose one predictor, and get the best fit you can.

4) Choose two predictors, and, using ozone as the response variable, fit

a regression "line".

a) How do you interpret the coefficients?

b) Is your model "good"?

c) Pick one of the two predictors -- it doesn't matter which -- and

replace it with another predictor. But here's the catch -- choose a replacement predictor that you think will be "better". How can/might you know which is better? Try it. Were you right? 5) Put all of the predictors into your model. Which are significant? Which

are not? What does this say about the relation of ozone to these

predictors? Would you be willing to believe that all of the insignificant

predictors really had no effect?

6) Notice in your last regression that temp was significant but temp2 was

not. But let's deliberately do something strange; remove temp from the

model and refit. Now what happened to temp2? Examine the relationship

between temp and temp2 and see if you have some idea why this happened.