What’s the difference between statisticians and data scientists in the context of the government statistical service (GSS)? There are differences in technical skills between the two but they can be bridged. The important difference might be more to do with the individual mindset of most data scientists and the organisational culture they seek – that might be the problem faced by the civil service when it comes to recruitment and retention.
Some of my recent posts, and the thrust of the Treasury’s Bean review, have alluded to skills needed in government to help bring the data offering up to speed. The GSS needs people who could assess and embrace the new worlds of open data, big data and admin data as old school statisticians are unlikely to do that on their own. Data science is a new and rapidly developing concept and means very different things to different people. This blog (“9 Steps to Become a Data Scientist from Scratch“) sets out what looks like a very reasonable list of the key skills needed to do data science (DS):
- Maths and statistics skills.
- Understand machine learning.
- Understand databases, data lakes and distributed storage.
- Data munging and data cleaning techniques.
- Basic data visualisation and reporting.
- Add more tools to your toolbox. (Think Hadoop, R and Spark.)
- Practice – or develop your own project, enter competitions, join a bootcamp, volunteer or intern.
- Become a part of the community.
At a high level, a description of the skills required of a statistician would not be out of line, according to those set out by the ASA. Yet, when it comes to the nitty gritty, not many of the DS skills would be required to make it as a statistician in government.
Those applying to join the fast track statistician stream of the civil service will read the outline. It does not set out the skills you need to apply but does refer to required qualifications: “You must have, or expect to get, at least a 2:1 degree in a numerate subject such as mathematics, economics, psychology or geography. You’re welcome to apply if your degree contains formal statistical training.” The “welcome to apply” if you have studied statistics does not sound like the technical bar has been set too high. There are various online job descriptions for statisticians and they too have only so much overlap with the nine points above.
By way of an example, The Economist recently advertised for a data scientist. Whatever people might think of the phrase “data scientist”, the job seemed pretty hard core. It’s a good comparison with the civil service as The Economist is dealing with main stream data and communication will be a core part of the job.
Many of the nine skills noted above appeared in the ad but it also had:
- Excellent communication and leadership skills
- An innovative mind set
- Excellent analytical and problem-solving skills
- The ability to work in a fast-paced and highly collaborative
- Good at building strong relationship across businesses as well as with remote global stakeholders of different cultures
- Committed and determined to deliver end goals
Sadly there are no statistician vacancies on the ONS job board at the moment that enable a fair comparison to see if the same personal qualities are needed.
I think that there is some overlap between a statistician and a data scientist but there are also differences. Perhaps the key issue is one not spelt out in the descriptions – that of the difference in mindset. The nature of a data scientist is innovative and disruptive – it’s usually about doing things better or faster or cheaper. It is often less constrained. While the civil service needs its fair share of that mindset, it is not yet common place in the public sector in the way it often is in the places where most data scientist are employed, such as start-ups and large profit-driven corporations. Getting the government data machine up to speed will need specialist recruitment but also a culture change. It won’t be easy and it’ll be harder to do it in Newport.