Big Data III – Degree or Not Degree? – Deep Thought Data Science

US comedian JP Sears weighs in [1] on the usefulness of tertiary education in the modern world – do the rising student loans and an oversupply of graduates still provide a competitive edge in the digital age of democratised information (like Wikipedia), search engines (like Google & Youtube), and MOOCs (Massive Online Open Courses)?

He’s not the only one asking these questions.

UK school leaver tertiary education rates rose from 19% in 1989 to 47% in 2016 but 1 in 3 can’t get a graduate job, according to a 2016 Guardian article on BIS government data [2].

Meanwhile in Australia Charis Chang’s article [3] on the OECD ‘Education at a Glance 2017’ report discovered,

“Once you take into account the cost of degrees as well as lost earnings during the years at uni, the report found Australian men got an 8 per cent financial benefit from going to uni. This comes from things like increased earning potential. But this was much lower than the OECD average of 13 per cent.”

This was despite the fact that, “At least 75 per cent of students benefited from public loans and scholarships/grants, such as the FEE-HELP program that sees students defer the cost of their education to when they start earning over a certain income.”

Things are arguably worse in the US, according to the Value Colleges 2018 article ‘Access and Affordability: USA vs the World’ [4]:

“In 1990, the U.S. ranked first in the world in four-year degrees among 25-34 year-olds; today, the U.S. ranks 12th.”

“The cost of higher education has surged more than 500 percent since 1985.”

“Today the United States has over $1.2 trillion of student loan debt with 7 million borrowers in default. … 4 in 10 millennials are overwhelmed by debt.”

Given the current crisis/problems with tertiary education and employment, its not surprising that more and more people are looking to the internet for training, whether self-taught or signing up to Massive Online Open Courses (MOOCs).

But what about data science specifically? Can you get by with some online courses or do you really need a Bachelors, Masters or PHD?

1. Yes

Burtch Works, an executive recruiting and market research company, puts it bluntly [5]:

“Data scientists are highly educated – 88% have at least a Master’s degree and 46% have PhDs – and while there are notable exceptions, a very strong educational background is usually required to develop the depth of knowledge necessary to be a data scientist.”

Rutgers Online has similar stats [6],

“If we look at the current crop of data specialists, we will see that nearly half of them have a PhD at 48%. A further 44% have earned their master’s degree [92% total] while only 8% have a bachelor’s degree.”

Data science communities and forums have been debating this for a while now. A KDnuggets discussion [7] had this to say,

“Your degree usually gets you in the door… you do not need a graduate degree to DO the job, but you might need one to GET the job.”

A data science reddit discussion [8] had similar sentiments,

“The only thing that is going to fill that void [of a degree] is actual work experience, and you’re back to your chicken and egg problem with getting the job, in order to build the experience.”

“… There simply is no substitute for the mathematics, if you really want to do data science. It is exponentially harder to do predictive work than it is to do descriptive statistics… I have a degree in math and I find it quite challenging to do good work.”

There’s a technical expertise in knowing the underlying theory and the intricate detail that is worth real money. Data scientists have a salary around $100k but the money isn’t just for analytics and software skills – data analysts have those and they only have an average salary around $60k [9].

Business doesn’t pay big salaries for nothing, they expect big returns. Data scientists don’t just need a good set of tools/software, with 50% knowing more than 10 distinct tools according to the 2014 O’Reilly ‘Data Science Salary Survey’ [10], they’re expected to build the best models with these tools to maximise efficiency, accuracy and speed.

So if you’re up for the challenge, where should you go for a good quality Masters degree in data science?

According to Data Science Central [11], the top 5 US Masters programs are

Harvard Data Science Course
UC Berkeley: Master of Information and Data Science (MIDS)
Stanford University: Master of Science in Statistics: Data Science
Carnegie Mellon University: Master of Information Systems Management
North Carolina State University: Master of Science in Analytics (MSA)

MSinAustralia.in ranked Australian universities’ Masters in Data Science based on ‘2016-2017 Times Higher Education Subject ranking for ‘Engineering & Technology’ [12]. They found

Monash University, Melbourne
University of New South Wales, Sydney
University of Melbourne, Melbourne
University of Queensland, Brisbane
University of Sydney, Sydney

SI-UK, a university consultant and student placement company, suggests the top 4 Data Science Masters courses in the UK as follows [13]

University of Stirling – MSc Big Data
University of Leeds – MSc Advanced Computer Science (Data Analytics)
University of Glasgow – MSc Data Science
University of Essex – MSc Big Data and Text Analytics

For similar university rankings across even more countries, see the 2016 article [14] by Alchemetrics , a UK marketing technology specialist company.

Some of these courses can be done entirely online, such as Master of Information and Data Science (MIDS) at UC Berkeley, USA or the Graduate Diploma of Data Science at Monash University, Australia.

Alternatively, many of the top US universities, including Harvard, Berkley and MIT, offer free online courses at edx.org , with the optional verified certificate for a small fee. While its not the same as a masters, its still excellent content from top institutions. Do you really need a Masters after all?

2. No

Content marketer Daniel Levine interviewed four professional data scientists in 2015 and was told [15]:

“You may not really need a degree in data science… More in support of self-learning than the damning of education, Edwin Chen explained, ‘A lot of the best data scientists I know come from fields that aren’t the fields normally associated with data science.’ ”

Mark Madsen continued this thread, “If you think about the early people who were doing [data science], they had a weird combination of math and programming and business problems. They all came from different areas. They grew themselves. The universities didn’t grow them.”

Further he says, “I have mixed feelings about the university programs. It seems to me that they’re more designed to capitalize on the fact that the demand is out there than they are in producing good data scientists.”

The author surmises, “There’s a tendency toward just repackaging existing courses into a coveted ‘data science’ degree.”

However, they do acknowledge some excellent university courses and the need for a level of training in theoretical statistics that is somewhat lacking in many data scientists right now, “Even at places like Google”.

Lastly they point out the importance of ‘soft skills’, particularly in communication, leadership, business and management. Randy Bartlett, “Criticizes university programs for often leaving these skills out all together: ‘There’s no real training about how to talk to clients, how to organize teams, or how to lead an analytics group.’ ”

The KDnuggets discussion mentioned above [7] also had some members who were skeptical of university courses and sympathetic to alternative learning options:

“There is a strong demand for analytic talent and a shortfall in supply. If you have a master’s degree, it will be add on for you but if you don’t have, many companies will overlook this as long as you have the right skills.”

“The current crop of Data Scientists has learned the work on the job. A lot of great research work is learned on the job. A Master’s or credential program could create problems for the person obtaining them once they get on the job and find that the work is as much experience as it is education”

“All the software/programming and the statistical/data mining techniques were not learned in college or a formal coaching environment. Most of the times, it was searching and reading on Google, a good discussion with the team, and in a few cases, a colleague or someone senior who will help when the right questions were asked.”

“To take a step by step approach an expensive Master’s degree may not be the best solution as no degree will eventually cover all these requisites and there will always be newer tech coming up. The best way would be to take a modular approach in learning all this stuff through short, inexpensive certificate courses. After all it is the knowledge which matters and certificate courses probably provide more hands-on, practical knowledge at a cheaper price than a Master’s.”

“Can you develop all these skills in one course? or a short program? NO! How about through a series of well-designed courses (not 1 or 2 perhaps 4 or more spread over a span of 1-2 years so you have time to assimilate knowledge and put those to work) that build on each other plus hands-on experience in working with complex data and models – Yes.”

Interested in online certificates and MOOCs now? Here’s a comprehensive list of them compiled by KDnuggets [16][17] and learndatasci.com [18].

But beware, with the growing popularity in data science, more and more dodgy half-baked courses are popping up, often repackaging 20^th century statistics and analytics as ‘data science’ instead of the ground-breaking innovations in theory and technology over the last decade.

Data Science Central has a great article, ‘Data Science Courses to Avoid’ [19]. So does Analytic Bridge, on ‘Fake Data Science’ [20].

David Venturi from DataCamp “Ranked every Intro to Data Science course on the internet, based on thousands of data points,” [21] and found the top three MOOCs for data science:

Data Science A-Z™: Real-Life Data Science Exercises Included (Kirill Eremenko/Udemy)
Intro to Data Analysis (Udacity)
Data Science Fundamentals (Big Data University)

He claims, “A year ago, I dropped out of one of the best computer science programs in Canada. I started creating my own data science master’s program using online resources. I realized that I could learn everything I needed through edX, Coursera, and Udacity instead. And I could learn it faster, more efficiently, and for a fraction of the cost.”

Meanwhile, at Data Science Central, Vincent Granville offers a 6 month ‘Data Science Apprenticeship [22][23]:

“For a cost well below $10,000, and with guaranteed job opportunities. The program would be open to everyone without screening, but the degree and the guaranteed jobs would be offered only to students with a successful completion of selected projects. If you don’t succeed, you don’t pay.”

Springboard offers a similar approach, with a 3-6 month intensive course and guaranteed employment or your money back [24]. A 2015 TechCrunch article [25] by John Russell explains,

“Springboard, an India-U.S. company formerly known as SlideRule, has raised a $1.7 million seed round to accelerate its concept of learning through engagement with others. It has adopted individual mentors, who provide a weekly catch-up session with their students, while partnering with other MOOCs for course content, creating its own where it sees gaps in the market.”

“The funding comes from some pretty prominent names, including LinkedIn cofounder Allen Blue, Princeton Review founder John Katzman, InMobi founder Naveen Tewari, and Wharton School professor Kartik Hosanagar.”

For an extensive review of the course itself, see [26]:

“The huge differences between a data science MOOC, like Coursera or edX, is that Springboard offers private mentor calls with real data scientists each week. In those calls, you can ask them anything, such as personalized feedback for your recently submitted assignments and capstones, what it’s like being a data scientist at their company, or anything you’re stuck on in your learning path.”

“Although the cost is a lot higher than a MOOC, it’s actually lower than every other online and offline data science bootcamp I could find (as of 2017). Compared to a university, the entire career track is priced very close to a single college course.”

“As of 2017, Springboard is the only online data science platform that offers a job placement guarantee.”

This is providing you are “Authorized to work in the US within 1 year following graduation from Springboard,” and are “Willing to live and work in one of eleven US metro areas”.

Conclusion

If you do graduate study, like a Graduate Certificate/Diploma, Masters or PHD, make sure it has a good balance of theory and practical problems, exposes you to cutting edge technologies and ideally has some business management thrown in for good measure.

If you’re learning online, choose excellent courses or programs even if they cost significant time and money. Make sure when you’re done your portfolio is full of real world data problems and solutions so that potential employers can see firsthand how useful you can be. Consider starting off in some kind of analyst position to develop some professional work experience.

Want more advice on how to kickstart your data science career? Check out the Kaggle CareerCon 2018, live streaming 20-22 March [27].