Jump to content
Urch Forums

SlowLearner38

2nd Level
  • Posts

    538
  • Joined

  • Days Won

    12

SlowLearner38 last won the day on February 9 2013

SlowLearner38 had the most liked content!

Converted

  • My Tests
    No

SlowLearner38's Achievements

Newbie

Newbie (1/14)

152

Reputation

  1. I'll reiterate what I always say in threads like this: Python. Python. Python. Learning Java from scratch for an economist is nuts. You're not building webapps with complicated user models, you're writing code to be used in research -- often nothing more than a simple script. Python is one of (if not the) easiest languages to learn. In fact, rarely do you need to have a deep understanding of Python to get something working. Sure C++ is marginally faster than off-the-shelf Python code, but for CPU-bound tasks, if you learn how to use standard Python libraries like multiprocessing (the goto method for doing parallel processing in Python), you can increase the speed of your code by a factor of N (where N == the number of cores on your CPU). Although I'm no longer in graduate school, I still do research with economics PhDs in academia and and IT consulting work for economics PhDs in industry, often helping them implement a programmatic solution to their problems. 90% of everything I do is in 1 of 2 language: Python or UNIX command line tools. Never once has my choice of language limited me. I have an open challenge for any economists to find a problem that they can do in another language that I can't replicate in Python. So far I'm batting 1.000. As others have said, you are an economist and not a computer scientist, which is the right attitude because you don't need to know the theory -- you just need to know how to use the tools to do the things you need to do. The best way to start down that path is to find a specific problem that you need solved and begin trying to implement it in Python. There are so many tutorials out there for Python it's incredible. Start there and learn as you go. Don't bother buying any books -- that's not a good approach for an economist to learn to program. Find tutorials online and try to replicate them, then tweak that code to your needs. Learning to program when you're not a computer scientist is best done with lots of hands on practice. Here is my blog where I have a handful of Python tutorials that you may or may not find useful: Python for Economists
  2. Hey hedgequant, I'm moving to Houston in July when the lab I work in moves down to its new home at Baylor College of Medicine and the CS & Applied Math departments at Rice. Looking forward to getting back to the south + not paying state income tax! I'd be happy to meet up once we're both settled in down there.
  3. I'm generally not a supporter of relying on others to collect and aggregate data for me, however, I recently came across Quandl - Intelligent Search for Numerical Data and it might be worth looking at. Basically, they have ways to access a variety of datasets from various sources using Python, R, MATLAB, Excel, etc.
  4. This is just a pet peeve of mine but the short name for Industrial Organization is not I/O, it's IO. The forward slash doesn't make sense. I/O is Input/Output. Just saying. :encouragement:
  5. If I would have instead picked Arizona around this time a few years back, I think I would probably still be in a PhD program. I agree with your assessment of their economics program. It also seems to be the program that you know the most about. Arizona is one of the few programs that doubles down on the microeconomic principles like game theory, IO, mechanism design, experimental/behavioral, networks and so on. In about 5 years from now, if you want a great paying economist or research scientist/associate type job at a tech company like Amazon, Google, Microsoft, Yahoo or eBay, then Arizona probably maximizes the probability of doing so. If you want to be a professor, then Arizona won't limit you from becoming one at a good school. Pick what ever school you want because at the end of the day it doesn't matter. The main influence on your eventual outcome is and will be you.
  6. Personally, I'd like to see Leontief input/output type models come back to fruition. [Then again, it'll be a cold day in hell before I ever step back into the realm of macroeconomics.] The idea that the macroeconomy can be reduced into a single model is insane. Heck, one has a hard enough time reducing measures like growth rates within a single industry into a 1 model.
  7. Wayback Machine to the rescue. Oldest record of the page is April 2009: FAQs : Doctoral Program : Department of Economics : Eller College of Management : The University of Arizona -- Verbal: 500, Quant: 770 Then the statistics changed in July 2011: FAQs : Doctoral Program : Department of Economics : Eller College of Management : The University of Arizona -- Verbal: 500, Quant: 780
  8. I've finally worked out a few bugs in a tool I've been building and I feel it's time to get some feedback. The tool is an ngram viewer/searcher for abstracts on SSRN -- here's the link: SSRN Ngram Viewer First of all, this is for searching unigrams and bigrams in the abstracts of research papers on SSRN over time and not the full text of the papers. Search queries can be filtered by academic institution as well as by the field which the paper was posted on SSRN (Economics, Financial Economics, Cognitive Science, etc). The point of the tool is to let people play around and spot trends across these categories, as well as help people find relevant research articles on SSRN. For a given query, click the dot on the line at a particular year to see a list of the research papers that make up the data for that point, along with the appropriate links to the abstract's page at SSRN to view the abstract and/or download a PDF of the paper. A couple disclaimers: It is against the Terms of Use of SSRN to automate the download of PDFs and I've done everything I can to respect all their policies. The website may go offline at anytime as I continue adding features and do maintenance. Respect the server or you'll be handled appropriately. On that note, feel free to share any interesting queries you come across or major bugs you see. This is still very much a "beta" version, so expect some bugs. Have fun!
  9. I'd rather call it privacy than anonymity, and privacy always boils down to economic trade offs. As it relates to this forum, I really don't care if people know who I am because everything I post is self-selected, which is the same rule of thumb I follow with any social network. It's a game that is solved using forward induction in my mind -- if I choose to release information that could jeopardizes my privacy in any way, then it must be that I feel the value added exceeds my loss of privacy. That being said, the Internet is written in ink, so playing the game improperly even once can result in potentially large losses. In general, I do feel that privacy is an extremely important topic for everyone. I remember in graduate school when I had a semester long debate with a professor about what the most important problem on the Internet is: I said privacy because it affects every single user of the Internet, he/she said something else. Even though I was confident I was right then, it's been nice to see evidence that economists are beginning to recognize the issue of privacy. For example, the most downloaded paper on SSRN right now is about privacy, and particularly why the "but I've got nothing to hide" argument is flawed. Also, if you've been following the FTC lately, you'll see that privacy is very high on their agenda (for example, in mobile apps). Information can be misrepresented, and the consequences of this can be huge. Think about it, let's say I apply for a million dollar life insurance policy. The insurance company will be more than happy to spend some money obtaining a profile about me that one of the data broker companies [which the public never hears about] has compiled. These black box profiles contain information aggregated across many avenues of technology that the Internet has made available. These black boxes are run by supervised algorithms that try to determine a person's type/preferences or whatever is relevant to their consumers -- marketing companies, insurance companies, HR divisions, etc. If I've learned one thing since moving my career more towards computation over the last few years, it's been the understanding of how 1 line of code on the backend of a website/app/etc, which the user will never see, can completely change their privacy experience. As it stands, these problems are going to persist for the foreseeable future for those who are uninformed, which is the majority (a digital divide problem). The fundamental problem is that the people who are doing the online tracking will always be more informed and have lower costs to become informed (economies of scale argument) than those being tracked. This is why I'm a supporter of movements like "Do Not Track" because the whole goal of it is to reduce the technical costs for those who care about privacy, but don't know how to actually protect it. The Do Not Track mechanism may not be perfect yet, but it's a step in the right direction. Sorry for the slight tangent, but I really believe these are important issues for everyone.
  10. Why is that unfortunate? Cleaning data properly often decides whether or not a project even gets off the ground. Being the data ninja in an academic environment like economics often leads to monopoly power and the thrill of gobbling up all the low hanging fruit.
  11. Good suggestion! There is a decent-sized learning curve for git, but it's definitely worth learning. The benefits of git are that it helps you keep track of changes you're making to scripts/files/etc that you are regularly working on, easier collaboration on these files and general increased productivity. If you're on a Linux/Ubuntu box, you can install git via apt-get by opening a terminal and entering: ~ $ sudo apt-get install git-core After installing git, cloning an existing repo (downloading it locally) is easy. For example, clone my public repo containing Python scripts for scraping data from newegg.com: ~ $ git clone git://github.com/econpy/newegg Moreover, here's the github documentation for creating a new repo: https://help.github.com/articles/create-a-repo. FWIW, students can obtain a paid account for free at github.com by following the directions here: https://github.com/edu. A paid account allows you to create private repos (free accounts can only create public repos). Fun fact about git: The creator of git is also the creator of the Linux kernel: Linus Torvalds.
  12. This thread seems fitting for a piece of motivational advice that I feel like sharing for all the new grad school hopefuls. It doesn't matter if you're in Economics, Sociology or Art History, the same logic applies: [video=youtube;STpd_VIZm-Y] The man, the legend. Bill F*cking Murray. The moral of the story is that if you know that what you're working on is important, then don't listen to anyone who tells you otherwise. At the end of the day, your biggest enemy in grad school will be yourself. Come to terms with who you are and what you want to do/change, because the sooner you do that, the better your plan of attack will end up being. "They can buy anything, buy that can't buy backbone."
  13. This may be among my top 5 favorite TM threads ever. I had the great pleasure of meeting Humanomics 2 months ago over some beers and grub in Chicago. Let me tell ya, that's a straight shooter with upper management written all over him. Well deserved! [clap]
×
×
  • Create New...