I arrived at the DTU campus on Wednesday of last week, a guest of Anders Gorm Pedersen. I reached out to Anders because he deals with Bayesian modelling and molecular data. The idea is that I spend a semester here and we write a paper together. Figuring out what the paper should be about is part of the work; we have an idea that we should explore approaches to the analysis of compositional data, which is relevant for the analysis of microbiome- and transcriptome data, among other things.
Compositional data are fractions and similar: data that sum up to a whole. Suppose there is a very big bag of blue and red marbles and we decide to draw five marbles from it. We do so and I tell you we drew three blue marbles; how many red marbles are there? It had better sum to five! In RNA sequencing you have the same type of structure but many more colors: you decide on how many reads you’re going to do beforehand so whatever you get out it has to sum to this chosen number of reads (or more likely some approximation to that number of reads). This sum constraint has various consequences for data analysis and interpretation.
Progress is modest so far! The first few days I mostly got admin and meetings done. Anders sent me a list of nearly 100 references, which I have started goring through with the goal of throwing away as many as humanly possible. Basically: look at title + abstract + maybe some figures, make a call on whether it will be useful or not. A lot of it does look useful, which is natural since it’s a curated list.
Getting to DTU from Hareskovby, where we live, is a bit of a hassle. You can go by train toward Copenhagen and back out and not get close enough. Or you can go first by train and then by bus, with 15 minutes wait between and then you’re on a bus. This option seems to take close to an hour door to door. In fact the best way (and this is very Danish) is by bike. This should take not much more than half an hour if you know the way. I tried to go by bike and spent close to an hour because I went the wrong way at least twice. But it has potential!
openalex.org is an alternative to Google Scholar. It looks pretty good to me, but I haven’t used it much yet. Their data is available through an API so you can for instance build your own citation graph. I might explore this at some point, perhaps when I have my shortlist of compositional data papers.
As far as I know Google Scholar is just one guy in a basement and we all remember what happened to Google Reader a mere 10 years ago.
Where have the websites gone? They are still there, but The Algorithm has made us bad at curation. This website in its current form (202312131102) is a contribution toward the curation that we stopped doing when the For You section appeared everywhere.
Some very nice visualizations about how small individual biases make large collective biases.
Snappy from Cory Doctorow:
Adding words to the plausible sentence generator doesn’t turn it into a superintelligence for the same reason that selectively breeding faster horses doesn’t lead to locomotives.
Backlinks: