Mon: Rapid post-publication review
Tues: He’s skeptical about Neuroskeptic’s skepticism
Wed: R sucks
Thurs: You’ll never guess how we answer this question: “Am I doing myself a disservice by being too idealistic in a corporate environment?”
Fri: Gresham’s Law of experimental methods
Sat: Turbulent Studies, Rocky Statistics: Publicational Consequences of Experiencing Inferential Instability
Sun: There are 6 ways to get fired from Johnson & Johnson: (1) theft, (2) sexual harassment, (3) running an experiment without a control group, (4) keeping a gambling addict away from the casino, (5) chapter 11 bankruptcy proceedings, and (6) not covering up records of side effects of a drug you’re marketing to kids
You put up a provocative title like “R sucks” and try to sneak it in as a December 23rd post? With vacations and holidays most people won’t get to it for another 2 weeks :(
Spoilers please!
Njnnja:
I’ll tell you this: The post is full of attitude.
Re Wednesday’s topic: So you would not recommend that I register for that R course on Coursera? (It’s either R or Python.)
Chris:
I find R to be very useful. R sucks but that doesn’t mean you shouldn’t learn it!
I need to learn both – at least to the point of having basic functional knowledge. It’s a question of which one I hit first.
Can you elaborate on your goals? Right tool for the right job.
> Can you elaborate on your goals?
Sure. The nature of my work is to feed data (in the form of bits) into a real-time or near-real-time processor which produces a high-level data product. I communicate with end users to establish what the high-level product is and, based on that and the characteristics of the input data, what algorithms need to run on the processor to turn raw data into data product. (To be more specific, input = multiple images and output = a single image annotated with a “Look over here!” cue when appropriate.) I do algorithm development in IDL and MATLAB. Both are good for prototyping. (This could be done in C or C++ but they’re not very efficient tools for that job.) I get a prototype processing chain worked out and then I provide a description document, code, sample inputs, outputs, and files with intermediate quantities to software engineers to create operational code (C++). IDL and MATLAB work fine for my purposes, however licenses cost a mint and (not unrelated to cost of licenses I imagine) people coming out school frequently have more experience using open source software for data processing, i.e., Python and/or R, than they do using MATLAB. (It is the rare undergrad or grad student who uses IDL.) To the extent that learning Python and/or R would enable me to get the most out of their skills, it makes sense for me to learn them. Also, and potentially more significant, perhaps the future of algorithm prototyping isn’t MATLAB but Python. NumPy looks pretty powerful and OpenCV-Python perhaps even more so. My sense is that I at least need to do my due diligence to establish if there’s a better way to do things than how I’ve been doing them for the past 15ish years. I’m a manager now and don’t do a lot of technical work but I need to be sufficiently aware of the state-of-the-art that I can establish that the way our group does our work makes sense – and change direction accordingly if it doesn’t.
Re Thursday’s topic: “No matter how cynical I get, it’s just never enough to keep up.” -Lily Tomlin
I am a R fan, so, I am excited to hear the opposing view on the matter.