2
$\begingroup$

I wish to break into the field of data science, more specifically, analysing data sets, extracting features and classifying them or identifying relationships.

However, I have no formal math background and am trying to re-learn it from the ground up. I've started an introductory stats and probability course on Coursera, and am looking at some linear algebra via Khan academy but looking at the math I'm still quite lost.

I was wondering if anyone knew of a good introduction to data analysis text that was aimed at non-mathematicians, a "black box" approach, if you will. That would allow me to learn some basic techniques while I progress in my math that I could be getting on with alongside my studies.

I've had a play around with MATLAB, sci-kit learn, scipy, numpy for Python but all the texts I use are very math heavy.

Is this unavoidable? Should I just knuckle down with the math first? What would your advice be?

$\endgroup$
  • 2
    $\begingroup$ Start with a problem that has data. It will contextualize the inevitable math. $\endgroup$ – generic_user Aug 13 '13 at 18:20
  • 1
    $\begingroup$ @ACD has good advice, but in terms of your question, virtually all the introductory statistics texts would claim to be aimed mostly at non-mathematicians. Big fat books like De Veaux, Velleman, Bock or Agresti and Franklin go some way into modelling. If you are not 20 or so any more, and find their style too juvenile, Freedman, Pisani, Purves is aimed at intelligent adults as well. $\endgroup$ – Nick Cox Aug 13 '13 at 18:26
  • $\begingroup$ Thanks very much for all your answers they are very helpful! $\endgroup$ – David Folksman Aug 13 '13 at 20:01
  • 3
    $\begingroup$ Possible duplicate of Websites to learn data analysis and visualization? $\endgroup$ – kjetil b halvorsen Nov 12 '17 at 15:08
1
$\begingroup$

Consider these Coursera courses:

https://www.coursera.org/course/compdata

https://www.coursera.org/course/dataanalysis

I don't have experience with them, but they look interesting.

Perhaps you could tell us more about your background so we can better recommend resources that fit your level.

$\endgroup$
  • 1
    $\begingroup$ I took the data analysis course with Jeff Leek. It was a good intro. It would be good if you had a little experience working with R or python (python isn't used but stuff you learned using python would be transferable). You will need to learn some stats, but you can do it, it just takes practice and a decent course. I took a Coursera course from Univ of Toronto with Prof Rosenthal which was quite good for people with no previous background. $\endgroup$ – chrisfs Aug 13 '13 at 18:40
  • 1
    $\begingroup$ I meant to say you'll need to learn some stats eventually, but not for the coursera course mentioned. $\endgroup$ – chrisfs Aug 13 '13 at 19:00
  • 1
    $\begingroup$ , in terms of background. I have just completed an MSc in Mobile Computing and come from a tech support background. I want to do a PhD in the future and I know I want to work with data. It feels like finally i've found a field that accomplishes real science in the field of computing! (as opposed to development) I'm eager to start so I came here. As I said though, my degree and masters did not include a great deal of math (if any). $\endgroup$ – David Folksman Aug 13 '13 at 20:02
  • 1
    $\begingroup$ @DavidFolksman: There are a lot of online resources for learning about specific topics, but it can be hard when you don't know where to start. Although classical methods are sometimes considered inferior to Bayesian methods, I really suggest starting with the basics – it may seem mundane but a firm grasp of elementary concepts is incredibly important for learning more advanced material. $\endgroup$ – Ellis Valentiner Aug 13 '13 at 20:32
  • 1
    $\begingroup$ THanks for the advice, do you have any advice on where to start? Whats the difference between the two? $\endgroup$ – David Folksman Aug 13 '13 at 20:51
1
$\begingroup$

Start with one of the contests on Kaggle. You don't have to win, but it will give you some data to work with and some room to experiment. @ACD is right that having a problem that needs solving involving data will give you context for the math.

As for the math itself, start with linear regression. Virtually regardless of the problem you're looking at, you can apply linear regression. It may be a wholly unsuitable approach, but you can do it and see the results and learn better ways as you go. You will have to keep referring back to the math as you are implementing the regression so you'll pick that up as you go along. As long as you know how to multiply matrices, you should be ok to start. There are other nuances like knowing what full-rank means, but that's one of those things you'll learn when you get an error message about aliasing.

And as for learning materials, I would recommend Andrew Ng's machine learning course on iTunesU. The problem with textbooks is that you'll spend all your time reading and not enough time actually doing. Once you get your feet wet, then textbooks are a good reference for learning new methods and getting more depth on methods you have learned.

$\endgroup$
-1
$\begingroup$

Personally i found that Using R and RStudio is the best for data analysis.

It's extremely easy to pick up, and very efficient. There's many packages using R that allow for some pretty cool visualisations.

Here's a link to the basics of R:

https://youtu.be/_lnDe3t97vo

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.