Here’s where my reading took me this week. Some of it is ecology, but a lot of it is other, random things that came to me via Twitter or from friends. Perhaps this will become a regular post for me!
Favourite paper Paine et al. on a wonderful idea of integrating demography and traits to predict community dynamics. I like the paper for its ideas – I have always enjoyed thinking about demography! – but I love it for its clear, lively prose. Like this sentence:
When ecologists observe phylogenetic community structure, they learn that something is going on, but it is rarely clear what that thing is.
… oh dang
I spent the early part of the week learning how to publish data to DataOne repositories from within R. I dutifully followed the EML documentation from the rOpenSci package EML.. then found I didn’t know what to do with my metadata once I had it! Fortunately I found the NCEAS Slack group, where Jeanette Clark was very helpful and directed me to this guide.
Questionable scientific practices
In the middle of the week, I read a fascinating preprint on the Open Science Framework, and then said this on twitter:
I just read this bombshell of a preprint from @OSFramework https://t.co/U3Id62b8Mm on Questionable Research Practices in Ecology and Evolution #ecopubs pic.twitter.com/kVRpiNSgCG— Andrew MacDonald (@polesasunder) March 21, 2018
and my mentions got.. a little busy. This is a topic ecologists are very excited to talk about, it seems. Many people expressed shock and outrage, but there were also many who expressed a weariness and sometimes, a fear for what this means for their careers and for our field.
I also learned a valuable lesson: if you tweet a paper try to find the authors first. This is basic and I should have known better! Thankfully Dr Hannah Fraser did eventually find the thread, so hopefully anyone who is interested to talk to her will be able to find her easily!
Reading Heather’s paper led me to the website for Transparency in Ecology and Evolution which looks like an amazing project!
As a coincidence, about the same time I found this podcast called Everything Hertz [https://soundcloud.com/everything-hertz/57-radical-transparency-with-rebecca-willen] wherein they interview a scientist who, among other things, outed herself and her colleagues for Questionable Research Practices. She just wrote out some True Confessions of p-hacking, HARKing etc. on her website. I tell you it is the most brazenly beautiful thing I have seen in so long; just looking at it is a source of catharsis.
Interesting StackOverflow post about why the confidence interval in linear regression is shaped like an hourglass
Kaggle has a blog, and they are doing these charming cartoon biographies called “Winding paths to data science”. Telling the stories of how different data scientists have found their calling.
My recent discussion on Twitter about whether PCA axes ought to be used as predictors generated a lot of comments! I’m very grateful for that. Several people recommended I start thinking about Partial Least Squares (“PLS”) which lead me to learn about Gaston Sanchez’s work
This beautiful piece about birds, and why they matter by Jonathan Franzen
I drew a lot of inspiration from “Lessons learned in Hell” written by “Phil” on Andrew Gelman’s blog. In it he describes the pain and effort behind a demanding statistical project, and one that he was doing based on other people’s results. This reminded me a lot of my own situation, and his lessons are useful for me and for all of us. Behold this immortal quote, one of the take-away rules he offers:
Try to find out all the ways your model is wrong.
- This disturbing story unfolding about Chris Wylie and Cambrige Analytica, covered by the Guardian. I had this open next to Andrew Gelman’s blog on “who will win the next election”. I have enormous respect for Gelman’s writing, but the contrast did strike me as very sharp. And mostly, the whole thing seems dark and twisted. On the one had we have political scientists who dutifully go to predict elections like they are natural phenomenon that just require the right sort of multilevel model. On the other, shadowy billionairs and naive gay data nerds are using data science to create weapons of disinformation for goverments to use on their own electorates.
My favourite RopenSci forum post is here, written by Noam Ross. In it he describes how to use worktrees so that you can version control something in a shared dropbox folder. This is something I do very often on projects I collaborate on!
One day when I live in the same place for an extended period of time, I will be able to invest in wall art once again, and I’m going to get this beautiful map of Canada with Indigenous place names
By an odd coincidence, Chelsea Little is trying the same thing (a post of her reading list) this week as well! go check out hers too!