Tuesday, November 06, 2007

That Painful Buzzing In Your Head? That's Cognitive Dissonance.

I just had a very interesting discussion with one of our local ronpaul "true believers." Flush with enthusiasm over Ron Paul's genuinely impressive $4-million-in-a-day fundraising success, this ronpaulista, he was also quick to brag that Ron Paul was recently polling at 7% in New Hampshire -- certainly an all-time high, and higher in that particular poll than both Mike Huckabee and Fred Thompson. I had the sense to declare it simply "a bump" -- they happen to all the candidates, after all, and besides, one data point does not a trend define. First, he responds "No way, man! He's been going up! It's linear!" Here he draws himself a little straight-line gesture in the air with his figure. "Actually, it's not even linear -- it's geometric!" (indicating, for you non-math-nerds, a rate of increase greater than a straight line). He then proceeds to accuse me of being "unscientific" for conjecturing that one single (relatively) high poll result might be, indeed, "a bump".

Of course, the first two comments didn't even remotely pass the "sniff-test", and that last comment durn near made my irony meter explode. So I decided to do the teensiest, tiniest bit of digging to provide a dose of reality to this otherwise affable fellow. Finding the conglomeration of recent polls at the indispensable RealClearPolitics, I have provided a nice, succinct plot for all our dear little ronpaulians who may be undergoing some undue enthusiasm as of late:

Now, I chose to put the y-axis scale at 100% just to gain some perspective there. But for those of you without huge monitors, the numbers at each vertex are 6, 2, 7, 2, and 1 percent, respectively.

Is that a "linear progression"? Not even a little.

How about "geometric"? Well, maybe since his peak, but geometric in the decreasing sense only.

How about "A bump"? Does that sound like a good description?

Unfortunately, much like the best weapons our military can provide used against invading martians, cold, hard data and logic have no effect on theronpaulexperience. If they did, I would expect to hear about 40,000 heads go *pop* as soon as I hit this "publish" button...

Update: Aw nuts. No popping.



Anonymous J to the D said...

You know, I just LOVE seeing political data plotted in what appears to be MATLAB. There's just something nifty about it.

Incidentally, does this ronpaulista of your acquaintance know the term "statistical outlier"? Just curious.

6:37 PM  
Blogger Benjamin said...

Heehee.. yeah, I figured you'd recognize the Matlabby-goodness.

I actually thought about updating the post by saying that I myself am experiencing some painful cognitive dissonance: Senor Ronpaulista is, in fact, a fully competent engineer with some decent technical chops, so when I take that fact along with his blind and foolhardy obedience to The Cause, it causes me an uncomfortable sensation, not unlike bad indigestion.

Incidentally, I sent him this graph, along with similar discussion to this blog post, and his response was basically to then claim that polls are inherently biased toward front-runners, and here's the biggest screamer: "I look at the St. Anselm poll as the only one that "captures" the true sentiment in New Hampshire" (the St. Anselm poll being the one that had him at 7%). So in other words,he takes the single largest statistical outlier and DECLARES that one to be the "best indicator" of the overall "sentiment".

Remind me to never let him do statistical work for me.

8:40 AM  

Post a Comment

<< Home