How I Became a Data Scientist Despite Having Been a Math Major(stiglerdiet.com)
stiglerdiet.com
How I Became a Data Scientist Despite Having Been a Math Major
http://stiglerdiet.com/blog/2015/May/11/how-i-became-a-data-scientist/
6 comments
Read as: how I became a data scientist after taking an appropriate and helpful major
The problem with not having a math background is that you will succumb to all the people who throw around buzzwords but don't know what the hell they're talking about. You need to be able to recognize (and assert authority over) someone whose first thought is that "support vector regression and a bayesian net" is the right way to go. Or, to put it nicer, refine their thought into something useful.
The problem with not having a software and "grungy coding skills" background is that you aren't able to efficiently reproduce/check/verify claims being made which are often wrong or misleading. A recent example I had of this is someone studying how an algorithm behaved on a subset of interest of a particular population, and they didn't even bother to check that the subset of their data was statistically large enough to draw any conclusions (only 17 records out of thousands). Needless to say this example also failed on the reproducibility and reuse fronts.
The problem with both in academia (even at polytechnic non research schools) is that they don't know enough about each other.
The problem with not having a software and "grungy coding skills" background is that you aren't able to efficiently reproduce/check/verify claims being made which are often wrong or misleading. A recent example I had of this is someone studying how an algorithm behaved on a subset of interest of a particular population, and they didn't even bother to check that the subset of their data was statistically large enough to draw any conclusions (only 17 records out of thousands). Needless to say this example also failed on the reproducibility and reuse fronts.
The problem with both in academia (even at polytechnic non research schools) is that they don't know enough about each other.
Question for the data scientists out there: is it more straightforward to become a data scientist from a software engineering background or from a math background? Data science is a varied field, but from what I've heard a lot of the work is munging data (which is a fairly easy task for a skilled software person).
Also, I'm glad the author added the twitter endorsement to the article --I've been considering creating a twitter account, and the fact that the author found it so useful helps with my decision to devote time to creating an account.
Also, I'm glad the author added the twitter endorsement to the article --I've been considering creating a twitter account, and the fact that the author found it so useful helps with my decision to devote time to creating an account.
The perfect data scientist -- imho is great at massaging data and has a ton of real life experience with real data; So somebody on the interface of statistics and computer science is best.
Computer scientists mostly don't know about sampling and make easy problems hard. Statisticians mostly fail at simply getting the data in a format they need.
Computer scientists mostly don't know about sampling and make easy problems hard. Statisticians mostly fail at simply getting the data in a format they need.
Perhaps PhD math programs don't hold your hand through teaching data science, but it isn't a horrible way of preparing you for a (first) job in data. Color me impressed when a history major picks up data science!
Yeah, I confess I only read the title, but I don't understand the "despite".
I mean, surely, you need a good background in mathematics to model and extract meaningful information from the data, so I don't see how it is crazy to do that with a PhD in mathematics ...
I mean, surely, you need a good background in mathematics to model and extract meaningful information from the data, so I don't see how it is crazy to do that with a PhD in mathematics ...
I think his point was that pure math can be very devolved from real world as I found when i tried to go from a Engineering to a HNC in Maths Stats and Computing.
Back then it was virtulay all Pure maths with the computing side being shall we say Antique - most of my course mates where working at Tier 1 RnD workng at the bleeding edge, my employer hired the first non academic knowledge engineer in the UK
Back then it was virtulay all Pure maths with the computing side being shall we say Antique - most of my course mates where working at Tier 1 RnD workng at the bleeding edge, my employer hired the first non academic knowledge engineer in the UK
That was my reaction too. I'm a computer science graduate who's constantly wishing he'd taken more mathematics courses so I could read and implement more ML algorithms...
I'm a math major and most of the expert data scientists I would recommend are social science graduates.
They have a lot more experience with real-world data and a lot more knowledge about what statistical techniques and assumptions correctly apply in real situations - quantitatively and qualitatively. You simply don't get this knowledge and experience from a maths-led statistics course.
I would be happy to recommend you a history major.
They have a lot more experience with real-world data and a lot more knowledge about what statistical techniques and assumptions correctly apply in real situations - quantitatively and qualitatively. You simply don't get this knowledge and experience from a maths-led statistics course.
I would be happy to recommend you a history major.
Yup, I was thinking that should read "...because I am a Math Major". Sounds like a natural choice to me.
As a math major working with data analysis and data science, I've come to the opinion that domain knowledge of the area you are analyzing as far far more important than being good at math. If I was hiring for a data science position in for example analyzing marketing data I'd definitely hire someone with a background in marketing over someone with a background in math all else being equal.
The highest GPA at my data science master course (Well, NLP and machine learning) was a history undergrad.
One of the best in my class was a philosophy major.
One of the best in my class was a philosophy major.
I went from sociology major to developer mostly on moocs. Minored math though.
Begin with Andrew Ng's mooc at coursera, to get the basics and the feel of it.
I've completed the very first his course in December 2011.
I've completed the very first his course in December 2011.