Drinking from a firehose

YES – I’m talking about grad school. I’m taking, what I understand to be, two of the most difficult courses to start with simultaneously in this program (many forums and the advisors themselves recommend you not do it). These are 6040 – computing for data analysis and 6501 Intro to analytics Fortunately, my computer science and extensive data analysis background is making 6040 a straight-forward endeavor. So far, this class takes me about two nights per week, and I’m about two weeks ahead, which means if I have to blow off a week, and I’ll get to that in a second, without any sort of negative impacts. Even then, that class, in particular, is a bizarre hybrid of basic introductory computer science and programming concepts mixed with fairly advanced and sophisticated data manipulation techniques. For me it’s a weekly ‘language drill’ – for the lack of anything else to call it. I am learning a lot, and the main thing I’m learning is that I’d be hard-pressed to justify the extensive use of python for data analysis in my current context. It’s fast and all, but not the kind of thing that lends itself to summary and computation of dozens of variables.

For the other class (6501), today is the last day I’m allowed to turn in my week 2 homework. This particular week’s lectures, notes, and homework took me, and I’m not exaggerating, just shy of 30 hours. Fortunately, the homework is done (only just, I realized today I’d neglected to scale the inputs for the model). Said homework write-up is just under 15 pages long (with tables, figures, text and shockingly little R code). This class basically gives you a concept, some cool R functions, a handful of sources, and a dedicated discussion forum (which is remarkably productive), and then the homework (or lab, whatever). The homework amounts to maybe 3 questions, one of which is meant to simply assess that you understand how to apply the method being discussed. The other two questions are brief paragraph asking for you to analyze some dataset using the concept & functions discussed along with some end-goal. Your job, as the student, is to analyze that data, write it up, and present it. Honestly, this is how I make my dollars. Granted, I don’t do the deep analysis, just the superficial numerical stuff, but this should be easy, right? NO. not even a little bit. Normally, I’ve got time, on the order of weeks to sift through data, and not only that, when I get stuck, I can set it aside for a day or two, work on something else and come back. This is not possible here. Not only are the concepts largely to me, I also have to develop an understanding well enough to explain the approach, methods, and outcome over the weekend.

Now, for the firehose part (bet you thought I was already there). The lectures and class discussion drop on Monday. This week, I had the good fortune of having Monday off, which afforded me the lovely space on Saturday to get into a totally unnecessary ditch involving a two dozen lines of code that was remedied Sunday morning with a single parameter… Anyhoo, There I was Monday morning, feverishly trying to figure out where to start with question 2 (of 3), when the new lectures dropped and the next weeks’ class discussion questions were posted. Plus, I knew (hadn’t checked, but knew), that this weeks’ 6040 homework had ALSO dropped. On top of all of this, I’m trying to get through a project at work that we do every other year, and helping to kick-off a massive annual project. The only saving grace thinking about next week is that I can reduce my effort in the other class washout falling behind there.

Monday, was not a great day. Ultimately, I kept my nose to the grindstone and things are fine. But, here I am on Wednesday, probably not able to take more than just these few minutes to myself to knock out a poorly written blog post before working my way through the lectures and notes for 6501 in prep for this week’s homework.

So that’s my update – I’m learning a lot, and not just like “oh, cool – I just found out about this k-means model in R, you can use for clustering”, I mean like: “Oh, so with k-means, I can probably develop a state-wide cluster-sample of communities to generate a reliable regional estimate of harvests – or at a minimum demonstrate that it’s not really practical.” Yes, I still have to learn a LOT with any of the topics we’ve been given, but it’s enough to actually implement some of the ideas in the real world – and that’s pretty cool.


Image by AntOne_01 from Pixabay

GA Tech vs. OU (which program?)

Wow. I haven’t blogged in a long time. To be fair, it’s been busy and the time I have had, I’ve chosen to spend on time relaxing. I’m only taking a few minute just now because the grad-school prep has kicked into high gear, and I have some thoughts on the process.

I wasn’t sure what I might expect in applying for grad programs. I did know that I needed an asynchronous online program delivered by a good and preferably recognizable school. A traditional education environment wouldn’t really work for me. After researching programs for a couple weeks, I narrowed the list down to 3 top pics: Georgia Tech (GA Tech), University of Oklahoma (OU), and Michigan Tech. I was also in touch with recruiters from three other schools that didn’t really fit my need. One of them appears to be more of a degree mill than the kind of program I want. There were two more I hadn’t quite gotten around to calling, but they were more on the expensive end and would be down the list. In any case, I put in for GA Tech first (my top choice), then OU (University of Oklahoma). I never got around to my #3 pick because OU accepted me before the ink was dry on my last letter of reference.

Many weeks after my initial application, GA Tech came through with an acceptance letter. Meaning I got into both schools and I had to make a choice. For me, it a huge surprise. Yes, I had really good reference letters, brought a pretty robust track record in data management and statistics, but I’ve been out of school for about 20 years. For OU, I was really confident about my application package. Given their acceptance rate, I reckoned I had a really solid shot. GA Tech, however was a different story. To start, it’s a notoriously difficult school to get into (23% acceptance rate?), and the Online Masters program is no different than on-campus. While I do have a good work-record, I was worried that my ‘pretty okay’ GPA and (while I think it was fantastic), relatively unremarkable university degree were going to work against me. In any case, I made the cut, and I’m feeling really good about that.

Anyhow, I got into both schools I applied to, so why did I pick GA Tech? That’s actually what I wanted to write this blog about. Perhaps you’re also looking at a data science degree and what to know why I picked one over the other? To be sure, I haven’t started courses yet so the relative value isn’t known, it’s just sort of anticipated, but I do expect to be challenged and learn a TON (that’s the point, right?).

The whole reason I want to get the MS is to expand my knowledge and skills in the realm of data science. Not just statistics or biometrics, but sort of overall. I’m a senior research analyst, but I actually don’t have the full background I’d need to go to another organization if I needed to (you never know, I’ve got no plans to leave, but I do have a powerful need to keep a paycheck). Plus, those skills can help us my current organization tackle research questions with even more capacity and knowledge. Just the other day, I was asked about a meta-analysis of income. This is a master’s thesis on it’s own, and it’s on my radar to just sort of slip it in at some point… How much better would it be if I took that database of well over a million rows and hundreds of variables and hit it with a giant statistical baseball bat? That’s the stuff grant proposals are made of. Anyhow, that was my frame of mind when looking at these.

OU (University of Oklahoma)

So, what did I think of OU? It looks like a ‘solid’ program. I can say that their student success game is good enough that I seriously considered just staying on track there. The program has an ‘average’, if slightly more competitive price-tag – About $32K USD in total (it was a bit more and I think subject to change, but it’s a university, that’s how it goes). The key for me, however, was the mix of coursework The classes I’d have taken are:

  • Required Classes:
    • Computing Structures
    • Algorithm Analysis
    • Advanced Analytics and Metaheuristics
    • Fundamentals of Engineering Statistical Analysis
    • Database Management Systems
    • Intelligent Data Analytics
    • Professional Practice
  • Electives (probably, it’s not clear what is offered in any given semester based on the materials I have)
    • Financial Engineering Analytics
    • Bayesian Statistics
    • Time series Analysis (I think? – this may not have been offered in a way I could do it)
    • Introduction to R
    • Advanced R
    • Data visualization

The total list of courses offered were fine for this kind of program, but missed some key elements, while involving things I’d like to swap out. Tor example the financial stuff and database management systems are low-priorities for me. Further, I don’t know that R would’ve been the best use of my time because that’s something I can learn on my own. Nevertheless, these remained the most valuable of the remaining courses.

Overall, the application and on-boarding experience was great, I was really impressed and never for a minute felt like I didn’t know what was coming. Thinking about that and the mix of classes, I’d describe This as the kind of program designed for people looking to transition into data and analytics from a related job or an entry level job to something more advanced.

This is a pretty luke-warm review of OU, but there is enough in there that It would have been a valuable learning experience. If you’re new to data science with some background or looking to jump up a level from junior or mid-level to a senior analyst, I think this is likely to be a really solid option and reasonably priced. And again, they have a VERY strong student success process.

Georgia Institute of Technology (Georgia Tech. / GT)

What was it about GA Tech? Well, if you’re researching these, you’ll probably have spotted that the price-tag clocks in at about 1/3 the usual price. I found it at $10K USD. It’s low enough, I even asked admissions to verify the price. That was the key attention getter and driver initially. Their admissions deadlines are a bit more stringent than other programs, and I found myself up against a deadline, so I applied before I’d fully digested other options.

I would describe the admissions process as ‘fine’, but a little slower on the information dissemination than I’d like, though it’s not proved a problem. I’ve received helpful responses to inquires in a timely fashion. One difference from OU was that this program also made it clear you need to be prepared with a strong STEM background to begin with, including pre-requisites such as Calculus, python programming, statistics and probability and Liner Algebra from previous schoolwork. This is also pretty typical among these programs, but OU placed less focus on that (Michigan Tech. my 3rd choice also emphasized these).

The course material, while still including things that aren’t essential for me, is very much in-line with my goals for this degree. Also, there are 3 tracks to help you focus your attention, I picked the Analytical Tools Track. The courses I’m going to try to take (again, not sure how scheduling works):

  • Required:
    • Introduction to Computing for Data Analytics (I may see about swapping this one out. The course description sort of suggests this is what I already do every day.)
    • Introduction to Analytics Modeling
    • Business Fundamentals for Analytics
    • Data Visualization Analytics
    • Data Analytics in Business
    • Applied Practicum
  • Electives
    • Regression Analysis
    • Bayesian Statistics
    • Probabilistic Models
    • Time Series Analysis
    • Computational Statistics

I think really, if you’re looking at something like this, unless you’ve already got an undergrad that’s strong in the maths or computer science, even getting accepted may be a challenge. It works for me, and I’ve got the background to launch in with an expectation of success. Term starts in about eight weeks, and I already have a significant amount of review and prep work to do (I’m about 3 chapters into my old Linear Algebra book at this point, and I’ve got lots more to do). I’m generally up to that challenge, as it is review, but I can see where some prospective students would be intimidated.

Other considerations: Neither program required GRE tests, OU is on a Fall-Winter-Summer schedule that appears to be partially aligned with the on-campus schedule, GT is fully aligned with a Fall-Winter schedule. Both programs are fully accredited and will result in a Master’s of Science in Data Science and Analytics. Both require a practicum, it appears OUs practicum comes from corporate partners or your current employer, where the GA tech one is less clear, seems to be either from your own job or they have ready-made projects that can be used.

So, that’s where I am. As I get into it and start really working toward this, I’ll be posting more about it, but for now, I’m going to unwind with some video games before bedtime.