Help me calculate expectation

Hacking employment, improving work, professional development
User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Help me calculate expectation

Post by 7Wannabe5 » Sat Jun 09, 2018 11:04 am

My MOOC provider claims that if I follow their suggested outline of courses towards expertise in the field of Data Science consisting of 400 to 800 hrs of study (give or take for prereqs in R and/or Python and variety of elective courses such as Econometrics.) The online courses in the collection were created at U of C Davis, U of M, John Hopkins, and a few other universities of similar caliber. This outline was created for me based on fact that I already possess a B.S in Mathematics. Approximately 25 years ago, I received scores of 770 in Logic and 790 in Math on GRE, so I theoretically should be able to accomplish this course, even given a bit of brain decay/decline.

My MOOC provider also claims that I should expect offer of employment with compensation approximating $103,000/year upon completion of course of study. I am hoping that those who are more knowledgeable about this market could verify this estimate and approximate likelihood and time until hire? I hope to retire completely by Harvest 2022, so I wish to compare the expectation of this course of action with other options. I should also note that the expense of this option will be only $59/month* on top of opportunity cost of time spent studying.

*So, seems too good to be true.

jacob
Site Admin
Posts: 10132
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: Help me calculate expectation

Post by jacob » Sat Jun 09, 2018 11:24 am

What are their placement rates?

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Sat Jun 09, 2018 11:32 am

@jacob:

I have no clue. That is one of the numbers I am hoping to get in the rough from fellow forum members. I should note that the course of study will be inclusive of a capstone project and an attempt to create a unique entry for a contest in the field. Also possibly relevant would be the facts that I am too old to care about age-ism and too obnoxious to care about sexism.

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Sat Jun 09, 2018 1:09 pm

First: What is the probability that you're going to complete all 800 hours of study? You DO tend to get sidetracked. Don't you?

User avatar
BRUTE
Posts: 3413
Joined: Sat Dec 26, 2015 5:20 pm

Re: Help me calculate expectation

Post by BRUTE » Sat Jun 09, 2018 1:45 pm

7Wannabe5 wrote:
Sat Jun 09, 2018 11:04 am
in R and/or Python
those 2 languages are extremely popular right now, due to absurd demand for Data Science. brute personally doesn't know shit about data science, but he's heard anecdotally that it's growing like crazy.

here's a popular survey from this year that lists Python, R, and several Data Science libraries as extremely popular:
https://insights.stackoverflow.com/surv ... technology
offer of employment with compensation approximating $103,000/year upon completion of course of study
for the HCOL urban centers, that's a pretty normal offer for recent grads (of both CS and boot camps). FANG will pay 50% more out of the gate.

brute isn't sure if those offers exist in the area of Detroit that 7Wannabe5 lives in. it is an unfortunate reality that these high-paying jobs mostly only exist in a few tech cities around the country (or even world): SF, NYC, Seattle, Austin. there are a few 2nd tier cities like Boston, LA, Atlanta, Chicago, Raleigh, and probably more that brute is forgetting. but this type of work is extremely concentrated in a few centers.

so it is possible that 7Wannabe5 would have to move to get such an offer. she should check out the Detroit tech market.

because Data Science is very new, shiny, and inter-disciplinary, brute has the feeling that it won't matter much that 7Wannabe5 took a course vs. having an official data science or CS degree. especially with the BS in mathematics.

another thing to watch out for is age discrimination. the tech industry believes that humans peak at 24 and should be taken out behind the wood shed at 35. humans with 3 years of experience in the industry are "Senior Engineers".

so 7Wannabe5 might have to fight a bit of a battle because she doesn't fit the typical dudebro stereotype that many companies are looking for. brute can honestly not speak to how much of a deal this really is (both gender and age), but has heard horror stories and "it's not actually that bad" stories on both issues. there are women in the tech industry, but they're rare. there are old humans, but they're rarer.

brute will say that the downside is just $59 * 12 * (2022-2018) =~ $2,850. plus, of course, the invested time. so if at any time 7Wannabe5 decides she wants to change course or that she can't do it (tech is not for all humans), she isn't out of $40,000. downside vs. upside on this deal definitely seem very good.

Glassdoor and other websites offer insights into actual salaries that employees at companies report, and the number of jobs in a city:
Data Science Jobs in Detroit, MI - 213 Jobs
Data Science Jobs in San Francisco, CA - 2,293 Jobs
San Francisco has only about 30% more residents than Detroit, yet over 10x the job offers for Data Science. the advertised salaries also typically start at 6 figures, whereas in Detroit, they start way below that (with some ranges going into 6 figures), though of course this has to be balanced against COL. there are no 1BRs for < $2,000/mo in San Francisco.

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Mon Jun 11, 2018 7:21 am

FBeyer wrote:First: What is the probability that you're going to complete all 800 hours of study? You DO tend to get sidetracked. Don't you?
Yes, but my primary focus towards which I keep redirecting my gaze and toes is 1)completion of permaculture project to point of "pop" production of approximately 1200 lbs. and 2) FI by Harvest 2022. I have constructed a complex web of fairly flexible sub-goals towards successful completion of these primary goals, but I am coming down to the wire on some branches that will require cash input. So, I either need to earn $40,000/year for two years, from 11/01/18 -11/01/20, OR $80,000 for one year within the that time frame.

I just entered into legal partnership with my permaculture-project buddy (who is also leader of community group), and we wrangled full approval from the city, so he will independently be completing vine fruit trellis, espalier, key-hole bed, asparagus bed, other projects that need a few years to "pop" this season. So, I can back-burner that primary project branch for a year and focus more on my financial branch.

Based on advice offered on this forum regarding influence of 5 people you hang out with the most, besides members of my family, my real life social circle now consists of only intelligent men who are quite affluent, frugal and/or avid gardeners (pick at least 2.) So, for instance, my BF is leaving for the Middle East to work a contract job at very high 90% tax free pay rate with housing/food included for free, and when he comes back he will have enough money to pay for his son's college and buy a top-line conversion van which he can live in (maybe with me.) Meanwhile, I will be camped out with one of his friends who has his own mini-orchard and is building a wooden boat in his garage. He is also an IT director for a large corporation, so he might be able to help me get a job locally. One of my ex-lovers with whom I am still friends is retiring from being director of IT for a large public entity this spring, so he would be another resource. My DD27 is now the most junior head-hunter for an esteemed firm that puts together IT teams, so she will be another resource. She told me that her firm attempts to convince employers to value online certificates, but a lot of them are still pretty old school, so I am putting that into the calculator.

@BRUTE:

Thanks for the overview. Very helpful. My DD27 is getting married next fall, so I wouldn't want to relocate until after the wedding. Another possibility might be what I could make if I joined my BF in the Middle East for a year at that juncture. I have family or at least friend-of-friend connections in NYC, SF, Austin, and Chicago, so some possibility I could finagle my way into the housing market on the cheap. Also, I might attempt my own van-conversion project while I am staying with my friend who has carpentry skills and wide variety of tools.

Anyways, I really should have put most of the above into a journal entry, but what I am hoping for is some degree of confirmation that my preferred path of study new skill set for a year/make decent money is reasonable choice vs. carry on for 2 more years making okay money. I should note that the financial gamble might be a bit more than just my time and nominal tuition because to meet this time frame, I will likely have to dip a bit into my savings to cover some living expenses, because all I will be doing will be studying at a computer and hiking in nature to recover from studying at a computer.

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Mon Jun 11, 2018 8:20 am

Ok you've got the connections in place at least. And you've got a carrot at the end of some arbitrary stick too.

It just sounds weird to me that someone with a BA in math can take an online course in Data Science and get a job... just like that. But again: connections and market demand might be an immense boon there.

For all intents I think, personally, that it makes sense, but industry tends to be conservative. AFAIK the best way in when you've taken an 'alternative' approach to education is to show off several projects you've done along the way, to show that you can in fact wrangle a crappy data set and produce insights from it. And most important of all: do you understand the needs of a business and can you COMMUNICATE those insights?

The 'data scientists' job[*] is to help those dealing with strategy make better decisions, the statistician's job is not to produce numbers. So keep that in mind while studying: How Does This Help Other People Make Better Decisions?

But uhm, if you already substitute teach, can't you teach full time and command a salary in the vicinity of 40K? I'm surprised that YOU of all people haven't found a way to weasel out some cash somewhere with the least bit of hassle... That's a complement by the way.

Post Scriptum: And... Congratulations on the legal partnership.
Also: get familiar with https://stats.stackexchange.com/

[*] Q: what's the difference between a statistician and a data scientist?
A: about 20K$ per year...

jacob
Site Admin
Posts: 10132
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: Help me calculate expectation

Post by jacob » Mon Jun 11, 2018 8:51 am

Q: What's the difference between a quant and a data scientist?
A: Financial skills

I looked up some data scientist requirements from job listing and they're basically quant requirements minus the finance. Insofar a MOOC is a fancier/virtual version of reading textbooks on your own, I might have some insights, since I'm also self-taught wrt the data mining, programming, and finance with only formal degrees in math/physics (+a bunch of physics experience that was largely completely irrelevant to finance). These [insights] would be null and void insofar finance is different from other sectors that employ data scientists. For example, finance might be more traditional/conservative in their hiring practices. All interviewees have to wear suits.

The biggest hurdle was in getting hired for the first job. That's why I asked about placement rates.

Students who go to Gilded U enjoy the benefit of job fairs, alumni connections, and intern-to-job recruitment (very popular since the company gets to try before they buy). "Top school" matters to the "business developers" (a fancy word for in-house recruiter) and recruiters because they often can't tell the difference between a matrix and a table and so have no way of assessing actual skill level or judging people by their own merits. As an outsider, it's therefore hard to convince them to give you a chance because you don't check the right alma mater box.

However, what I've seen working for others (who didn't enjoy the benefits of the Gilded U track) and what worked for me was to build a useful product and use that as an entry ticket. In my case it was a complete backtester that processed tick data for trading along with a strategy that ran on it. Of course it wasn't nearly as sophisticated as the one they used, but OTOH in the dozen or so candidates I eventually interviewed, none of them actually walked in the door and showed me a job-relevant project they had constructed themselves. At best it was some tangentially relevant school project and even that was rare.

https://en.wikipedia.org/wiki/Kaggle (*) used to be great for sourcing data to work on. Maybe it still is. People doing well in those competitions got hired. Maybe they still do. Some of the competitions looked a lot like slave-labor though... kinda like when a company hires out their work as competitions so they get 10 finished projects but only has to pay the best of them. Awful practice. However, one can use such competitions to help getting a real job. If they're still around I suggest getting some experience in there.

I recall one (small) company which instead of a job listing asking for a resume had a coding problem instead and in order to apply one had to send in a solution. Caveat emptor --- they might be trying to solve their own problems for free like some did in the competitions above.

(*) Don't make the mistake of presenting their training project as some clever/quirky insight/interest you had. Someone actually did that to me. It worked pretty well for them too (got past the recruiter, etc.) until they ran into me and I happened to know where this strange and unusual interest in Titanic survivor rates came from :P ... That was pretty bad :lol:

There are also data science bootcamps: See https://www.switchup.org/research/best- ... -bootcamps

I've written a recommendation for a guy to go to one such bootcamp (was required for him to get accepted). IIRC, it was #6 on the list.

Finally, unless you're a superstar, it's really important to live in the same city as the one you're trying to get hired in. Companies just don't want to bother with long-distance hirings when they got 500 locals knocking on the door already. One thing that might work is to arrange some informal interviews with multiple companies while visiting a city. People love to talk shop insofar one makes it easy for them (e.g. offer to buy lunch or coffee).

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Mon Jun 11, 2018 12:05 pm

FBeyer wrote:It just sounds weird to me that someone with a BA in math can take an online course in Data Science and get a job... just like that.
Yeah, me too, thus this thread. OTOH, there is the possibility that the fact that I took this shiny, new approach to education might make me seem more youthful to potential employers? I mean, interviewing for a job with a 28 year old when you are 53 can't be any worse than getting naked in front of one, right?
But uhm, if you already substitute teach, can't you teach full time and command a salary in the vicinity of 40K? I'm surprised that YOU of all people haven't found a way to weasel out some cash somewhere with the least bit of hassle... That's a complement by the way.
Thanks. I am weaseling free housing out of my social circle, because there is nothing else currently available for less than 10% of my current income. Substitute teaching only pays around $15/hr. Private tutoring pays around $20/hr. Seasonal work at gardening center pays around $11/hr. So, I am currently making over $40,000/year, but only because I am working more than 40 hrs. I am also having to eat a lot of transition costs such as Ubering across town from one job to another. So, the cash flow problem is nominally solved, but not in a manner that is in full alignment with my preferences.

@jacob:

Gotcha. My learning path would include a micro-course which is entitled something like "How to Win a Kaggle Contest" and a capstone project.
(*) Don't make the mistake of presenting their training project as some clever/quirky insight/interest you had.
So, if my 3 initial project ideas were:

1) Something to do with northward earthworm migration after global climate change apocalypse.

2) Design of Arbitrage Weapon to be aimed at the Empire of my arch-nemesis J.Bozo.

3) Maximization of employee utilization towards profit matrix at uber-big-box gardening center based on analysis such as is conventionally used for detection of individual involved in cashier theft being applied to detection of which employees are better at selling $1400 lawn tractors to old men vs. those who are better suited for lifting 30 bags of mulch on to a cart.

Your advice would be that I should go with (3)?

User avatar
BRUTE
Posts: 3413
Joined: Sat Dec 26, 2015 5:20 pm

Re: Help me calculate expectation

Post by BRUTE » Mon Jun 11, 2018 12:15 pm

jacob wrote:
Mon Jun 11, 2018 8:51 am
"Top school" matters to the "business developers" (a fancy word for in-house recruiter) and recruiters because they often can't tell the difference between a matrix and a table and so have no way of assessing actual skill level or judging people by their own merits.

..

what worked for me was to build a useful product and use that as an entry ticket.
this is the case for almost all tech recruiting brute has seen. not only are recruiters and HR humans incapable of judging developers before hiring, so are other developers. in fact, it's difficult to judge developers one has worked with for quite a while.

thus, reputation and track record become vitally important. for developers, track record equals projects published on GitHub that are somewhat relevant to the industry/domain. or at least show that 7Wannabe5 can roughly do what she says she can. apparently 90% of humans who apply for tech jobs can't even do the simplest FizzBuzz test.
jacob wrote:
Mon Jun 11, 2018 8:51 am
unusual interest in Titanic survivor rates
haha, brute has worked with that demo dataset too :)

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Mon Jun 11, 2018 12:30 pm

What's wrong with presenting a Titanic-like data set? (Also: that's the one I learned logistic regression on :) )
As long as you're up front about why and in what context you made some analysis, I really don't see the issue.
It's always better to Show than Tell.

To get paid you don't have to be original, you have to be good at what you do.

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Mon Jun 11, 2018 12:51 pm

@BRUTE:

One of the problems with being an old female human is that until fairly recently, I would have had no idea what GitHub was, and I still had no idea what FizzBuzz was until I just looked it up. But now that I looked it up I am more cheerful, because I could have passed that test in 1982 typing BASIC into my cassette tape-driven VIC 20. Super easy, but also super boring compared to writing a Gypsy Fortune Teller Program.

I forgot to mention that I will be done with the lost-lab pre-req I assigned myself within the next few weeks. The Google Certified Entry-Level Information Technology Support Specialist coursework is so easy for me I can keep up with it even when I am working 60+ hours/week at my 3 jobs. I have determined that I don't want to be anything like a full-time Systems Administrator, because that job involves routine and maintenance, and I am terrible at both of those things. However, it will be interesting to see if I get any inquiries after my resume is auto-distributed when I am awarded this certificate.

jacob
Site Admin
Posts: 10132
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: Help me calculate expectation

Post by jacob » Tue Jun 12, 2018 7:20 am

Haha, I crashed and burned when I first ran into the fizzbuzz test in a phone interview. It's pretty far removed from numerical/scientific programming which admittedly is not the same as enterprise level programming; so there's the risk of testing a baseball player's throwing ability by asking them [randomly] to throw with their left arm. I don't think it's a great test, because it has a bit of a "are you smarter than a 5th grader"-vibe to it---like please show how to calculate 5*8 using new math :geek:

Another crazy stunt that used to be common in the quant world (perhaps it still is, I hope not) was using sudoku and brainteasers as a proxy for IQ tests (which are illegal). The first headhunter I talked to back in 2008 specifically told me to spend 25% of my time learning C++, another 25% on SDEs(*), another 25% on finance, and the last 25% on .... brainteasers. Brainteasers are usually easy to explain if you have some time to think about it or you've heard it before, like fizzbuzz, and somewhat random(*) if you're given 30 seconds to work out a solution on the spot.

(*) Which would be the hard part of the actual job.

Showing a project or past work that's relevant to the company at hand is crucial. Like brute said, for a software developer, you show your work on github and that project best be somewhat related. IIRC, Tyler9000 used to show a widely sold cellphone cover he had designed (as an industrial designer for design jobs). It's an effective approach.

(*) Unless, we're talking about the meta-test of BS'ing one's way through an exam---something that's probably relevant to sales.

When I first started pursuing quant jobs, I naively sent in my academic CV listing all the papers I had published, all the talks I had given, etc. That CV never even got the attention of recruiters. When I changed the CV listing all my papers about some result to a resume featuring single bullet point about a subresult saying something like "Increased accuracy over existing models by 16%" (even though that was not the crux of the science), it was like flipping a switch. Suddenly the recruiters "understood".

If you want to do science on earthworm migration, you better tag some additional work onto it showing the financial/economical impacts to the fertilizer industry. Similarly, if you're analyzing traffic patterns on google satellite, ... turn your attention into stock/flow analysis of the parking lots of competing companies (location, revenue, ...) if you want to work for a corporation; or look at smog or green space if you want to work for the city. Outside of academia, all scientific/nerd work has to have some kind of money or power connection.

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Wed Jun 13, 2018 7:01 am

@jacob:

Well, part of the reason why FizzBuzz challenge would be super-easy for me is that I often play the Buzz game with very young children. So, I have recently had reason to recall the notion of Mod function. Playing Buzz with 5 year old children is like playing Peek-a-Boo with a baby who has not yet developed concept of object permanence, because even the very brightest children can't predict the outcome. OTOH, there are some conversations about math that I can have with a very bright little 6 year old girl from Bangladesh and a very bright little 7 year old boy from Yemen which I can't have with the majority of the other teachers in the teacher's lounge, because the children are more open-minded and curious.

Anyways, that's why whether or not any of my studies ever lead to lucrative employment opportunities, I think continuing to throw down approximately $50/month on whatever bright, shiny MOOCS or books strike my fancy is a good investment.

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Fri Jun 15, 2018 3:28 am

Guy lays out an approx. 6 months roadmap for how to learn Data Science.
https://www.youtube.com/watch?v=MOdlp1d0PNA
  1. Learn the math (2-3 months)
  2. pick a language and learn the fundamentals of the language before you dig into the ML aspects of the language (1 month)
  3. find data sets and ML tutorials and do them (1-2 months)
  4. do a passion project (1+ months)
Track everything on GitHub so you have a live repo you can show off to employers. Make certain you visualize and attach a narrative to all your findings so you get used to producing a story, not just numbers.

With my background I have to agree. This is the order in which to do things. Especially the fundamentals of the programming language before the specific ways to implement ML in that language.

User avatar
7Wannabe5
Posts: 3771
Joined: Fri Oct 18, 2013 9:03 am

Re: Help me calculate expectation

Post by 7Wannabe5 » Fri Jun 15, 2018 6:55 am

@FBeyer:

Thanks. Very good resource. The path recommended for me by my MOOC provider is similar, but based on assumption that I already completed (1.) This is true, but that was almost 30 years ago. So, might be good idea for me to 1) Review the math (1 month) before I begin either the Python or R batch? (I took a Fortran course in 1985, so maybe I should choose R ? :lol:) Since it is pretty easy to "pause the timer" when doing online coursework, I was thinking I would just review the math as necessary.

jacob
Site Admin
Posts: 10132
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: Help me calculate expectation

Post by jacob » Fri Jun 15, 2018 8:07 am

Python and R complement each other. Fortran can substitute for python.

R is for drawing statistical conclusions, whereas python is for creating data with computations that R is too slow or lacks nuance for. Think of R as a statistically loaded version of Mathematica or Matlab. It's easy to write things up, but you don't want to wait hours or days for certain calculations that fortran et al would do in 10 seconds. I never bothered to learn python, since I just got my heavy numbers out of my java or fortran programs.

People just learn python because it's a good starter language that's also useful, kinda like Fortran and Pascal which is what I was subjected to in the 1990s.

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Fri Jun 15, 2018 2:25 pm

FØRTRAN can substitute for Python? What exactly do you mean, 'cause last time I checked, those two languages we're pretty much polar opposites.

Anyone reading this: if you're new to programming, FORTRAN is probably the worst first language you could pick up to begin with, perhaps with the exception of Brainfuck. FORTRAN is useful later, much later, iff you work in a very numerically heavy environment (ie physics, bioinformatics, and/or theoretical chemistry).


FORTRAN as a starter language... :lol:

User avatar
FBeyer
Posts: 912
Joined: Tue Oct 27, 2015 3:25 am

Re: Help me calculate expectation

Post by FBeyer » Fri Jun 15, 2018 2:49 pm

7Wannabe5 wrote:
Fri Jun 15, 2018 6:55 am
...So, might be good idea for me to 1) Review the math (1 month) before I begin either the Python or R batch? (I took a Fortran course in 1985, so maybe I should choose R...
If you've completed a fortran course, at any point in your life, you'll pick up python in no time. Don't worry. Do you have a link to the course curriculum you're looking at?

For didactic reasons it is naturally better to refresh your math ASAP. Then you'll forget a bit again once the course starts and you'll refresh it once more, with less hassle and with better long-term retention.

ML is an odd branch of mathematics, in that it is very visual, compared to what I'm used to. You can draw a lot of the things you're doing even without a great depth of mathematical understanding. In linear regression, say, you'd normally have to develop a rather high abstraction level before you realize that your design matrix is simply an idempotent operator, which just means that you project you data points from a possibly infinite space, into your chosen space, and then measure the orthogonal deviation to estimate the best fit. Once you get the projection part of regression, you can start making drawings of what regression looks like in multiple dimensions. ML is often the exact opposite. You can start with a drawing, and then you can tack all the what-the-fuck-does-that-mean-math somewhere on the drawing. The more your ML drawings make sense, the more certain you are your math is correct. In math, people who can draw the ideas are usually wizards, in ML almost everyone starts with a drawing. It's very comforting in that regard :D Also your pen 'n paper work looks much cooler!

Edit: Also. Be aware that once you're 'done' and out and working, you'll probably just be doing a metric fuck ton of logistic regression anyway, so get that down pat! Including all the other link functions and all the ways of comparing model accuracies BIC, AIC etc and really know all the ins and outs of LogReg. And Neural networks of course.

jacob
Site Admin
Posts: 10132
Joined: Fri Jun 28, 2013 8:38 pm
Location: USA, Zone 5b, Koppen Dfa, Elev. 620ft, Walkscore 73
Contact:

Re: Help me calculate expectation

Post by jacob » Fri Jun 15, 2018 3:24 pm

Data science probably IS numerically intensive, no?

I never took a programming course for any language ... or maybe it's more fair to say that I never learned anything I didn't already know from the couple of mandatory "programming for physicist" I had to take. In the 1980s/1990s everybody started with some variant of BASIC. I picked up my first programming book in the 4th grade. Fortran77 very much looks like BASIC with a few quirks (mostly relating to input and output). Fortran90 is a bit more advanced.

The biggest shift for me was switching to OO languages. It took me a couple of years before I grokked it.

I think the advantage with functionally oriented languages is that you can hack up something much more quickly than if you have to start figuring out how to how to organize your objects and classes. It's possible to define, set up, and multiply two matrices in fortran90 using TWO lines of code. Try doing that in java/c++ ... I don't know how it's done in python.

Anyhoo... my point was that R and python are complements. It's not possible to pick one or the other w/o "hardship". Also not sure I would categorize R as an actual programming language. Sure, one can do loop and conditional statements, ... but the point of R is really to help doing high-level math.

Learning fortran these days is probably only relevant insofar you work in one of the hard sciences and need to understand some old code from 1965 or 1992 or whatever.

Post Reply