Combining some mechanistic knowledge and Occam's razor appears to me to be a very powerful tool for getting closer to the truth! Let me show an example that I believe is related to this, which I found while writing my last published paper. It's on the underlying temperature-dependence of diapause 🤓❄️ - ThreadSky

vonschmalensee.bsky.social • 84 days ago

Combining some mechanistic knowledge and Occam's razor appears to me to be a very powerful tool for getting closer to the truth! Let me show an example that I believe is related to this, which I found while writing my last published paper. It's on the underlying temperature-dependence of diapause 🤓❄️

Comments

vonschmalensee.bsky.social•84 days ago

Quick background: we did a pretty laborious experiment to empirically determine thermal reaction norms for diapause termination in a butterfly. The result was the two curves below. Also, we found some signs of a third reaction norm for diapause induction that might look something like the red line.

vonschmalensee.bsky.social•84 days ago

Rewind to 1997. Three researchers publish a paper on the same topic, with the figure below (https://doi.org/10.2307/2404844). Very reminiscent of our results. However, this paper has remained in relative obscurity, only racking up 15 citations to date, despite being on a quite popular topic. Why?

vonschmalensee.bsky.social•84 days ago

Well, they don’t estimate the curves directly. They just observe insect emergence in a variable thermal environment, and iteratively tweak curve parameters in search of the best explanatory model, under some constrains (the main one being how the shape of the curve can vary, see figure).

vonschmalensee.bsky.social•84 days ago

As a result, the curves from the previous figure emerge. The three distinct phases are just a consequence of picking the sparsest solution to the problem (see the image). This (Occam’s razor) combined with the constraints on the reaction norm—the mechanistic insight—got close to the truth!

vonschmalensee.bsky.social•84 days ago

I think biologists are often sceptical towards this approach. The mantra “correlation does not imply causation” is deeply rooted in many of us. Perhaps this is why this paper by Johnsen et al. has remained under-cited for decades. But I think there is great power in approaches like this!

vonschmalensee.bsky.social•84 days ago

In signal processing, methods to reconstruct under-sampled signals by promoting sparsity has been used for quite a while now. And this has led to some very impressive decompression methods, where signal is seemingly squeezed partly out of thin air (check this: https://web.archive.org/web/20160120072320/http://www.technologyreview.com/view/515651/bell-labs-invents-lensless-camera/)

Comments

Posting Rules

Reply