Recovering from the red-eye, this was a low-research day. Tarmo Äijö (SCDA) gave a nice talk in the morning about modeling biome populations in the human gut, using multinomial distributions with a time-dependent set of population fractions, which themselves come from a Gaussian Process. He uses STAN to do the sampling; it is impressive the complexity of the models that STAN can handle, and sample well. They find changes in the biome during and after diseases, and also some species that seem to show periodic population changes, with long periods (many weeks). All of the data are from genome sequencing, by the way: It is impossible (apparently) to tell apart these bacteria on morphological grounds.
Late in the day I had a call with Kelle Cruz (CUNY) and Ellie Schwab (CUNY) about low-mass flaring stars. They have a nice data set with many observations of many stars, and most of them are non-detections. They have the upper limit data, but not the measurements in the non-detection cases. They want to say things about the distribution of activity, as a function of stellar type. We tried to hash out the simplest possible project, and then decided to defer to a face-to-face on Monday.
No comments:
Post a Comment