Proof Our Readers Are Smarter Than Us?
Speaking of my “Tom Carr bust,” check out the comments on my Tom Carr post from yesterday (which provides links to the whole sordid affair).
Tom Carr and I (and Dominic Holden, and others) are locked in a debate over the statistical significance of certain numbers on Seattle’s marijuana filings. Well, turns out that if you raise the issue of statistical significance on the Slog, you get comments like this:
So, now, if you do a linear regression on the model: 1/(# of filings) = A + B*year + C*(0 if before I-75, 1 if after) + error; you find that it fits the data really well. The model is statistically significant at levels as low as 0.001013% (5% or 1% are usually considered convincing), and it explains around 98% of the variability of the data (adjusted R-squared = 0.9859). Further, the estimated value of “C” is significant at levels as low as 0.012%, and the value of “C” is positive, confirming that years after I-75 are correlated with lower anual filings (or higher 1/filings). The estimated value for “B” is significant at levels as low as 0.62%. Linear regression assumes that the “error” values are independant, identically distributed, normal random variables. Checking the “residual” errors of the fitted model various plots indicates that these assumptions have not been violated.Conclusion: At a 5% (or even 1%) significance level, there is a statistically significant correlation between the passage of I-75 and reduced annual marijuana filings. (But remember kids, correlation does not imply causation).
Thanks for posting, “student.” And thanks to you too, “j-lon,” for your thoughts. Now, can someone, in a very remedial way, explain in the comments section of this post what the hell “student” just did?
God Damn. I want to go on a date with that person. They can lecture me, blackboard, erasers and all, anytime they want.