I keep talking about using large datasets and how some models take quite long time to run. One of those models is typically two way fixed effects models. While regular econometrics packages like plm are quite good for reasonably large datasets, more and more I find it hard to use with my datasets. Here is where LFE comes in. Developed by Dr. Simen Gaure, this package demeans the data in two dimensions (typically the demeaning only happens in one dimension).
The result is an extremely fast estimation for even large datasets like mine. I could say I fell in love with it the moment I saw it. The one problem I was having with it was to figure out how to run some instrument variable diagnostics on it (Sargan test of overidentification).
Yeah, I know. Open source, inspect the source code, develop your solution… I don’t know if I ever told you how much I hated debugging other people’s code? Well for me there was an easier way. I reached out to Dr. Gaure and he responded within days with a solution. Turns out the code was ready to ship with the next version of the package. I won’t share the code here in case Dr. Gaure wants to post it himself, in his own channels.
This to me is another reason why Open Source is the way to go. While I -like many others- did not make use of the source code. I benefited from the mentality behind open source in solving my problems. I think open source aligns perfectly with what we try to accomplish as researchers.