Next time I have to teach the difference between explanation and significance, this will be my go-to example: “Scientific aptitude better explains poor responses to teaching of evolution than psychological conflicts” by Mead et al. in Nature Ecology & Evolution.
The article has several figures that look like this:
The P-value looks amazing (9 x 10^-15), but the spread is equally large. I would not even want to guess what the R-square of that relationship is.
… but still is specific to each R script file. For instance, a lot of R users in our group load their packages at the top of their R scripts. Often, these packages are the same, but sometimes they add a different one for a specific type of analysis. I use the tidyverse all the time, and in some analyses vegan. A school of thought would be to automatically add these package to the start up using a .
This is something we have known for the last 10 years (or 20) through for instance the research done by Eric Mazur and collaborators.They showed that using peer instruction improved student learning of physics concepts by 20% in a real-world, class setting, compared to a “regular” lecture-based approach with the sage on the stage imparting their wisdom to the young susceptible minds!
That is a fine example of a very deliberate teaching practice.
I am currently teaching a graduate “Stats” course, which is more a historical exploration of statistical issues in ecology, led by grad students. As part of the course, we are also exploring best practices in R and ecological data management. So naturally we covered Brian McGill’s 10 commandments for good data management, and his follow-up post with an example application of these recommendations with a toy data set.
I decided afterwards to do the challenge, and with our weekly University of Guelph R Users group (UGRU) we walked through the code line by line, and discussed why certain lines were included, alternative ways to code them, advantages and disadvantages of these alternative approaches.
Talk about climate change, and be sure that your analyses are rock solid, because you will get some serious backlash. What interests me most is the danger of “if the only tool in your toolbox is a hammer, everything will look like a nail” thinking.
http://www.statisticsblog.com/2012/12/the-surprisingly-weak-case-for-global-warming/ is a blog post written by a graduate statistics student, with this summary:
“TL;DR (scientific version): Based solely on year-over-year changes in surface temperatures, the net increase since 1881 is fully explainable as a non-independent random walk with no trend.
I sometimes think of myself as a science nerd, but then I read Benjamin Mako Hill’s computing set-up, and suddenly it’s like reading the uber science nerd manifesto. Amazingly consistent and hard-core, every single step of his work flow. And where our work flows overlap is probably in R:
“R is slow and using it with big datasets gives one plenty of time to reflect on this fact. But R is also expressive, elegant, and concise for numerical and statistical work so I happily suffer through it.
An interesting paper in Science by Stumpf and Porter takes a hard look at “general” power laws in science:
“A striking feature that has attracted considerable attention is the apparent ubiquity of power-law relationships in empirical data. However, although power laws have been reported in areas ranging from finance and molecular biology to geophysics and the Internet, the data are typically insufficient and the mechanistic insights are almost always too limited for the identification of power-law behavior to be scientifically useful (see the figure).
Who thought stats would enter into every aspect of life? What about a Belgian dj duo called Mumbai Science, with the track “Ancova”?
And this thanks to 2 Many Dj’s and Music for Life 2011.