Thread by @WomenInStat

One more quick tweet, unrelated to the Gelman-Rubin diagnostic.

Someone asked, "I hear C++ is fast but a little hard to grasp. That true?"

Mostly yes. Like Python, R is mostly easier to learn and often is slower than C/C++.

I recommend you think about how your code will be used when you decide what language to code in. If you're coding for yourself and you probably just need to run it once, then R may be a good choice. Optimizing for speed may be overkill. (2/)

If you are writing a function/package for public consumption, then speed is much more of a concern. You can profile your code to see which parts are time-consuming. You can also just google what things R is slow at (ex loops). (3/)

For ex, in my R package glmm, I wrote my code initially in R because it's easy to write and debug. Then I rewrote the computationally expensive parts (ie pretty much everything once you pop the hood) in C. (4/)

I did this because glmm uses Monte Carlo (which is inherently computationally expensive). (5/)

I translated function-by-function into C, starting at the most basic building blocks. With each translation, I checked to make sure the calculations were identical between R and C. (6/)

In case you didn't know, you can call C from R. The way the R pkg glmm does this is with the .C function.

If you're into C++, you can check out the R package Rcpp, which helps you call C++ from R.

#Rstats (7/)

But remember there's a trade-off. R is generally faster/easier for a human to code in while C/C++/Fortran are faster for the computer to beep-boop through.

You don't get extra credit for coding everything in the fastest way possible. You can optimize your code... (8/)

Or you can optimize your LIFE by considering both the time coding/working and the time you spend socializing, exercising, relaxing, etc. (9/)

With that, I'm seriously out of here. Time to go live the other parts of my life. Take care and thanks for reading!

Cheers!

(10/10)