Very cool project and possibility to work 3 years developing Stan and collaborating with me (Aki) and other Stan development team. Deadline for applications is 22 October.
Quantitative Modelling of Cell Metabolism (QMCM) group headed by Professor Lars Keld Nielsen at DTU, Copenhagen, is looking for experienced Bayesian statistician for a postdoc position.
Group specializes in creating complex composite mechanistic / data-driven models of cellular metabolism.
Such accurate and interpretable models can be used to analyze response to drugs, search for a new pharmaceutical targets or explore basic biology.
Some of responsibilities include working with well-known Stan framework: improving it for HPC usage and fitting large nonlinear models.
This kind of activities supposed to be carried out with collaborators in Aalto University (Aki Vehtari) and core Stan developers at Columbia University.
All details could be found here
Wow! So cool.
I’m trying to understand why p-hacking and HARKing are not acceptable, as “data-driven” methods of analysis, while other post hoc methods of fishing for correlations are considered very respectable and forward-looking. I’ve asked some researchers, e.g. @Kordinglab on Twitter, but I’ve yet to get a clear answer. Can anyone explain?
Lydia:
Exploration is great, and hypothesizing after results are known is fine too. The mistake that sometimes arises is when people then attribute too much certainty to their speculations. An example would be finding a statistically significant comparison from data and then acting as if this represents a general pattern in the world. It’s fine to speculate, but it’s a scientific error to (a) act too certain in this speculation and then to (b) dismiss later evidence that is in contradiction to earlier, overconfident claims.