Careers in Data Science

What do data scientists do and how do they do it?

1 July 2022

About me


  • BSc(Hons) Mathematics
  • MRes Statistics and Operational Research
  • PhD Statistics and Operational Research
  • Data Scientist @ Jumping Rivers


A typical week


  • Data science consultancy
  • Internal projects
  • Training courses
  • Outreach work e.g. conferences

Projects

Dashboard development


  • Developing dashboards to display data for monitoring
  • Dealing with data updates
  • Maintenance of dashboards
  • Deploying dashboards


Blog

Statistical modelling


  • Statistical advice
  • Reviewing potential methods
  • Building models
  • Scoping projects


FAO and NU Solve

Programming and implementation


  • Writing code to implement models
  • Reviewing code
  • Upgrading from e.g. Excel


Skills

Problem definition




Data visualisation


Visualise your data at the start…



… and at the end.


Data visualisation


Initial data exploration:

  • Usually standard plots
  • Check model assumptions
  • Look at relationships between variables
  • Why not summary statistics?

Data visualisation


dataset mean_x mean_y std_dev_x std_dev_y corr_x_y
Data 1 54.26873 47.83082 16.76924 26.93573 -0.0685864
Data 2 54.26588 47.83150 16.76885 26.93861 -0.0686092
Data 3 54.26785 47.83590 16.76676 26.93610 -0.0689797
Data 4 54.26327 47.83225 16.76514 26.93540 -0.0644719

Data visualisation


Data: {datasauRus}

Data visualisation


Equations look scary…



… plots not so much.


Data visualisation


Use annotations


Data visualisation


Use annotations


Colour sparingly

Data visualisation


Use annotations


Colour sparingly


Keep it simple


Statistical methods


  • Hypothesis testing
  • Time series analysis
  • Monitoring
  • Modelling and predicting

Statistical methods


  • Start simple
  • Awareness of methods (not necessarily experience)
  • If you know a method, you also need to know it’s assumptions

Biggest differences in industry


  • General vs tailored solutions
  • Explainability sometimes matters as much as statistical performance
  • Faster deadlines

Questions?


Nicola Rennie