Thursday, May 31, 2012

Hypothesis Testing

Web of Science has a great service that notifies me when a paper I've authored has been cited.  This is a great way to find out who is reading -- well, at least who is citing -- one's work.  But it came as a bit of a surprise to find our 2008 paper on parent-aneurysm angle cited in a paper entitled "The Role of Hypothesis in Biomechanical Research".  Hmm, I thought, that can't be good...

The bottom line is that the authors looked at 100 papers published in The Journal of Experimental Biology and the Journal of Biomechanics and observed that, while the majority purported to test a hypothesis, many actually did not.  The authors go one speculate about a number of possible motives for such "window dressing".

I was somewhat relieved to find that our paper was only "strongly suspected" of having such a presentational hypothesis. Unfortunately the authors did not pick our paper as one of the four cases that they used to justify their suspicions, but they did later highlight the kinds of things that gave them pause:
  1. In a small number of cases, the data did not appear to bear on one or more of the hypotheses that it was the stated intention to test.
  2. In the other cases, which were more common, one of the stated aims of the paper was to test a hypothesis that was widely accepted to be true or false
  3. A sub-set of these cases were hypotheses that we thought to be post hoc; typically, in such cases, we registered only a suspicion.
I thought our paper was very clear about its motivations, namely that after we had observed "distinct types of intra-aneurismal hemodymamics" in two patient-specific cases we'd previously studied, we "hypothesized that these two distinct 'hemodynamic phenotypes' were primarily the result of the angle which the parent artery makes relative to the nominal center of the aneurysm bulb, independent of the bulb shape itself".  We then stated that we would "test this hypothesis using an idealized basilar tip aneurysm model in which [this angle] could be controlled independently," and did just that.

In the Discussion we did say "it can be hypothesized that Type II flow would create a higher associated risk of rupture", but this was intended to be conditional -- perhaps we should have said "could" instead of "can" -- but we made no claims about testing this.

So I'm not really sure why those authors were suspicious of our motives. Our data, from an idealized aneurysm model with varying angles, obviously bore on our hypothesis.  I'm pretty sure our hypothesis was novel, but even if it wasn't it certainly was not widely accepted to be true or false. Finally, we did not go through the trouble of running careful simulations on a series of idealized models with progressively angled bulbs only to hypothesize about it after the fact.

Oh well, as Oscar Wilde, er, hypothesized, the only thing worse than being talked about is not being talked about.

1 comment:

  1. FYI, Google scholar offers a similar service. Log in to your g-acct, go to scholar, and click "My citations" (top right). You can even create a public profile page from that info.

    Mine is