Last time | Next time |
Connor received a gold start for asking about when to stop our modeling: why not consider a cubic, or a quartic, or a quintic, or....
We tried a cubic with the Keeling data, and checked the confidence interval for the cubic term, and it was significant at the 95% level. Interesting!
We should realize, however, that we're only looking at bi-annual averages over some smallish part of the Keeling data. When I went back to my office, I used the entire Keeling data set, and this is the result of the Mathematica command "ParameterTable" (which gives t-statistics and P-values, rather than confidence intervals -- I hope that you're comfortable with both!):
Alexander also illustrates that one cannot simply invert the regression equation $y=a+bx$ to get the regression equation $x=\frac{y-a}{b}$. So it really matters in linear regression which variable is considered "independent" and which is considered "dependent".