The sharing of data is one of the key principles of reproducible research (the other one being code sharing). Using the data and code a researcher has used to generate a finding, other researchers can reproduce those findings and examine the process that lead to them. Reproducibility is critical for transparency, so that others can verify the process, and for speeding up knowledge transfer. But recent events have gotten me thinking more about the data sharing aspect of reproducibility and whether it is tenable in the long run.
Software has for while now played a weird an uncomfortable role in the academic statistics world. When I first started out (circa 2000), I think developing software was considered “nice”, but for the most part, was not considered valuable as an academic contribution in the statistical universe. People were generally happy to use the software and extol its virtues, but when it came to evaluating a person’s scholarship, software usually ranked somewhere near the bottom of the list, after papers, grants, and maybe even JSM contributed talks.
I recently finished reading Steve Coll’s book Directorate S, which is a chronicle of the U.S. war in Afghanistan post 9-11. It’s a good book, and one line stuck out for me as I thought it had relevance for data analysis. In one chapter, Coll writes about Lieutenant Colonel John Loftis, who helped run a training program for U.S. military officials who were preparing to go serve in Afghanistan. In reference to Afghan society, he says, “Everything over there is about relationships.
Back in February, I gave a talk at the Walter and Eliza Hall Research Institute in Melbourne titled “Lessons in Disaster: What Can We Learn from Data Analysis Failures?” This talk was quite different from talks that I usually give on computing or environmental health and I’m guessing it probably showed. It was nevertheless a fun talk and I got to talk about space-related stuff. If you want to hear some discussion of the development of this talk, you can listen to Episode 53 of Not So Standard Deviations.
Several times over the last few weeks my hatred of Doodle polls has come up in meetings. I think the polling technology is great, but I’m still frustrated by the polls. Someone asked what I’d rather have happen and I said: “set the meeting, then let me know when it is, if I can come I will. if i’m not there then i’m happy for you to decide without me”
Defining success in data analysis has eluded me for quite some time now. About two years ago I tried to explore this question in my Dean’s Lecture, but ultimately I think I missed the mark. In that talk I tried to identify standards (I called them “aesthetics”) by which we could universally evaluate the quality of a data analysis and tried to make an analogy with music theory. It was a fun talk, in part because I got to play the end of Charles Ives’ Second Symphony.
The NIH is soliciting input for their Strategic Plan for Data Science. If you are interested, today, April 2, is the deadline. You can provide input here. Below is what I plan to submit. Summary My main critique is that the report is somewhat vague. More specifics and concrete examples should be included. My main concern is that the draft describes initiatives with the goal of improving the back end of data science (data storage, data management, and computing infrastructure) without realizing that to do this one needs to understand the needs of those working on the front end of data science (data exploration, quality assessment, interactive data analysis, and method development).
Anil Dash asked people what their favorite file format was. David Robinson replied: CSV is similar to Markdown. No one global standard (though there are attempts) but a damn good attempt at "Whatever humans think it is at a glance, they're probably right" — David Robinson (@drob) February 8, 2018 His tweet reminded me a lot of this tweet from Stephen Turner In defense of Fahrenheit pic.twitter.com/qwDcBm0XVr — Stephen Turner (@strnr) February 20, 2015 There is a spectrum for tools from the theortically optimal to the most human usable.
In this post I describe the dslabs package, which contains some datasets that I use in my data science courses. A much discussed topic in stats education is that computing should play a more prominent role in the curriculum. I strongly agree, but I think the main improvement will come from bringing applications to the forefront and mimicking, as best as possible, the challenges applied statisticians face in real life. I therefore try to avoid using widely used toy examples, such as the mtcars dataset, when I teach data science.
Editor’s note: For the last few years I have made a list of awesome things that other people did (2016,2015, 2014, 2013). Like in previous years I’m making a list, again right off the top of my head. If you know of some, you should make your own list or add it to the comments! I have also avoided talking about stuff I worked on or that people here at Hopkins are doing because this post is supposed to be about other people’s awesome stuff.