Tuesday, November 25, 2008

Correlation, Causation, and Circularity

I have noticed two main trends in my life since starting grad school: drinking and debt. Data shows that these two factors are trending in the same direction with time (see Graph), leading to the question of whether this is causation or mere correlation of trends. The first point of significance is that both plots are trending in the positive direction with respect to time, meaning that my amount of debt and alcohol intake are both at all time highs this very instant. I lift my glass to that. The second point of significance is that both drinking and debt values show decreases at the same points in time followed by roughly parallel spikes following the relative lows (Point A and B on the graph). Unfortunately the bin size of the plot is too large to detect whether a lag is apparent between an increase in drinking and an increase in debt. However, raw data indicate that the changes in total debt lags slightly behind changes in total alcohol consumption, indicating causality. Multiple factors such as deadlines for conference presentations, preparation for comprehensive exams, chapters of your dissertation due to your advisor, grant applications, and hospital stints for nervous breakdowns can lead to brief declines in alcohol consumption, resulting in fewer happy hour expenditures. And, following the physical laws of the Universe, these brief "dry spells" are always followed by extreme drinking and debt spikes, like fern spores in the rock record after a mass extinction.

And so here is another night I didn't spend working on my dissertation - guess I'll just finish this bottle and take another semester to finish to my degree. With friends like Jack Daniels, who needs a Ph.D. and bank account in the black?

Hufflebug said...

A most fascinating analysis! However, I would be curious to know how the potential confounders of "attending conferences" and "darn my computer just crashed AGAIN" could be accounted for. After all, both of these could independently increase drinking and debt. Additionally, "attending conferences" can increase debt before the rise in drinking due to up front costs such as booking flights and hotels months before said conferences. . . which would rather put a kink in the debt-lagging-drinking hypothesis. Similarly, when the expenses are subsequently reimbursed 4 months later, that may decrease debt while significantly increasing drinking. Which, per the most insightful Diagram provided, will lead to increased debt.

Furthermore, the "darn my computer just crashed AGAIN" factor can either increase or decrease debt before the drinking commences. It can increase debt by means of having to pay the department back for throwing a several-thousand-dollar-worth computer setup out the window; however, the time lag it would take the department bureaucracy to charge you could cause the debt to actually follow the drinking that would presumably begin shortly after said act of throwing computer, etc, out the window.

On the flip side, the "darn my computer just crashed AGAIN" factor also has the potential to decrease debt if one chose to head butt the defunct computer, causing a concussion that may be eligible for workman's comp. Additionally, given the sympathy said concussion would garner, friends may be guilt tripped into bringing baked goods and other delectable treats, therefore decreasing one's grocery expenses and helping to pare down the debt even more.

So while this study is an excellent start, clearly, further research in this critical field is needed.

In the interest of expediting this endeavour, please excuse me while I go head butt my computer.

PS: I prefer milk chocolate.

Laura said...

Well, the actual conference attendance was factored into the debt and drinking spike after the preparation for conference presentation dry spell. However, you are correct that computer crashes were not factored into the analysis. It will have to be factor into subsequent analysis. This is what peer-review is all about...

But if you're going to drink chocolate milk in my presence, make sure its in happy coffee or a White (or rather brownish) Russian.