They also made the same lecture only using Julia rather than Python. I personally found it to be really useful and I taught myself enough Julia to be a danger to myself and others.
Just to point out: the co-author is Thomas Sargent, Nobel Prize winner and generally a big deal. If he's taught himself Python, then kudos (he's 76). Even if not, it's a strong recommendation to have his name on the cover.
I don't think pedantry about the name is a useful contribution to the conversation.
While it's true that Economic Sciences prize is not a "real" Nobel prize, it is commonly referred to as a Nobel prize. Interestingly, the Nobel Foundation also lists "Economic Sciences" on their website listing Nobel prizes even though they do not award or fund it: https://www.nobelprize.org/prizes/
There is no conversation here. I was merely taking the opportunity to point out that there is a common misconception regarding the "Nobel prize" and the Nobel Memorial Prize in Economic Sciences. Two distinct prizes, one commonly mislabeled. Repeating the misnomer just normalizes the error.
Um, yes, there is a conversation here. It's the comments on this thread. And, until you hijacked it, the name of the prize was not what we were talking about.
Yes, I made it a topic as the term was used incorrectly. So I was pointing this out, as not to further this misconception. Is there any additional discussion on this topic needed?
If you're interested in econometrics, I highly recommend checking out Marc Bellemare's "Metrics Mondays" blog posts, which are full of useful, pragmatic advice for applying econometric methods to real-world data: http://marcfbellemare.com/wordpress/metrics-mondays.
If you're coming from an ML-focused approach to statistics, studying econometrics can be an interesting change of pace, because the focus is totally different. ML practicioners tend to be focused on prediction, while econometricians tend to focus on causal inference - utilizing pseudo-experimental variation within the data to estimate causal effects between variables. This turns out to be really hard to do correctly, and learning the pitfalls can make it easy to identify potential weaknesses in other research.
Most econometric work has historically been done in Stata, although it seems like both R and Python have been increasing in prominence a bit recently.
When I was in school around 2010 or so, a lot of the younger econ grad students were primarily interested in R. I don't think Stata's going away any time soon, but it might not be completely dominant for that much longer.
A lot of people I know at various departments are switching their undergrad stats/econometrics classes from Stata to R. That's the beginning of the end of Stata.
That matters, but I don't think that's happening until all of the big graduate-level metrics textbooks get R versions. And even then, at least a few papers are going to run into trouble with older reviewers who are used to seeing work done in Stata and don't trust anything else.
Yes, and it's also non-trivial to write R code that matches your textbook's answer if your textbook used Stata. You have to do things like look up which specific variant of the sandwich estimator Stata uses for robust standard errors, so you can tell R to match that.
In Stata's defense: It helps that Stata is actually really good for the "running regressions" part. In particular, it gets robust standard errors right without much extra work in complex cases that would require a lot of additional code in Python or R.
R wins easily for data visualization and scripting, though. It's also much better as a skill you can "take with you". If you end up working in industry, you may not be able to expense a Stata license, but you'll almost certainly be able to use R (although maybe not RStudio).
You don't even want to expense a Stata license. Stata is the worst thing I've ever had to use. Maybe as a person who can't program it makes sense, but as a professional developer almost everything about Stata is non-intuitive, confusing, and stupid. Also the only thing to go on is their stupid pdf manual. Finding real people on the internet who actually use it is almost impossible.
Ahh, this is nice. I work in an office with a number of economists (energy economics consulting firm), but I’m basically the only python user. Feel like this could be useful in bridging some gaps for the folks who only use SAS and got their PhDs cobbling together whatever code (VB, FOTRAN, etc.) got their models to run back in the day.
I’ve written so much documentation on Confluence where it would have been easier to just send a pdf like this :/
I want to learn Julia but I have a very big concern: does it actually alter your personality in a way that makes you condescend to everyone about their inferior programming languages, or is it just that people who already are condescending choose to learn Julia?
I remember thinking about this before I knew JULIA. I can't remember that time clearly. I have tried to black it out. Presumably, I was just sitting nude in a cave bashing two rocks together covered in faeces and confused shame...just like you.
Here are things I can guarantee: learning JULIA will make you stronger, more agile, your IQ will double, women will be able to smell your dominance, children will run from you screaming in terror, you will be able to grow a thick lustrous beard (even if you are a woman), you will be able to talk to animals and lead them in battle, and you will be able to throw a spear through a 5m deep concrete wall from 200m.
EDIT: I forgot, if you do learn JULIA be sure to avoid any contact with indigenous societies. They will likely think you are a God. You go to the Amazon one time, and suddenly these people are building shrines, making human sacrifices, and carving intricate wood etchings of benchmarks and terse, readable function compositions (they told me they were still using Python2.7...lol).
LOL. That's a shame. I like children, so I guess I'll just have to stay locked in this naively blissful void that I've been mischaracterizing as a 'brain' my whole life.
Answering your question in good faith, even though I am unsure it was asked that way-
I am not a Julia programmer, I mostly write in python, but I find their community welcoming and not condescending at all. I think it would have a positive impact on most people’s personality
The language is very interesting too but doesn’t yet have a google, apple or msft behind it so I would understand why lovers of it maybe overstep a little promoting to try to keep it alive
Personally I find the integration with cuda to be really well done and I could see it being easier than python for highly customized deep learning (custom kernels etc)
Your comment above seems kind of unnecessarily mean spirited to me - maybe I’m reading it wrong?
I was surprised - because I remember you responding to the “I made 500k with machine learning guy” and being really impressed with your willingness to try to teach the guy without shitting on him (I’m an ex algo/hft guy and think someone with your knowledge could have gone that route very easily)
It is unjustifiably mean spirited to Julia programmers in general. I admit that it is an unfair characterization of everyone who programs in trendy languages. I just got irritated that he commented with pure snark to what looks like an amazing paper. And I find being witty and mean instead of blandly authoritarian is the best way to handle those people.
Anyone who wants to learn, great. Anyone who wants a one-sentence snark, I'm not going to be as open to helping out.
There is no need for generalization, there are many people in the community that respect and enjoy other languages, and most people also frequently use Python and R for most things that Julia is still not mature enough.
But it's certainly hard sometimes for people who learned of powerful non mainstream languages, having to see people putting an amazing amount of resources and effort to provide every functionality to mainstream less powerful languages that would be almost free in said powerful language (be it syntax extensions with macros, high performance dynamic code without using FFI, parallelism, better compile-time checking...). It's probably what Lisp users had to deal with for 60 years now. Or more recently people who learned Rust but still have to deal with a world of C++.
Quick answer: Julia is often faster than Python and other high-level languages. Rather than writing high-level code in Python, R, or Matlab and performance-critical code in C, the idea is that one writes the whole thing in Julia.
Edit: I agree that the "...more interesting" comment above sounds condescending. I have not found the Julia community to be condescending.
One of the thing I like from Julia compare to Python is that it have the concept of missing data representation. Python's data science library represent it via NaN or Null which is good enough for most cases but not all cases.
I often wish R's syntax was cleaner and faster, Julia is may accomplish this. I don't think Python is a great substitute for R in many areas where statistic is heavily used and influenced. I've used Python for Deep Learning and NLP. Time series and many other statistical base stuff I use R.
It basically assumes you have at least one year's grad school level background in economics. It does give some overview, but probably not enough to learn it from the book alone.
I would recommend 3blue1brown only if you've already covered the material in another way. It's a great way to get some new intuition about things, the videos can help something 'click' and are a pleasant watch with an obviously high production quality. I just don't think they are great for being your first exposure to a topic.
https://lectures.quantecon.org/jl/