Abstract: Statistical modelling is an error-prone endeavour. Mistakes are easy to make and hard to detect. For over a decade now, Michael Brand has been running regular peer reviews for data science projects, and almost without exception these reviews brought to light serious issues that required major revision to the analysis.
In this interactive tutorial aimed at data scientists of all verticals and experience levels, Michael steps through some of his own real world past reviews, demonstrating to participants how to run such analytics peer reviews on their own, and the range of blunders that they can expect to uncover.
Beyond reviewing, individual data scientists will learn how to avoid the common cognitive traps in their own modelling work, while leaders of analytics teams will gain insights on how to root out bad practices and shift to more reliable, predictably-good data science work methods.
The tutorial will be split into a number of case studies, each case study walking through the review process of a single model. Each review is a puzzle, and participants will be invited to put on their thinking caps, find the clues, and crack the riddle of where analysis mistakes were made and how to correct them. Each case study will highlight different types of errors and different types of tell-tale signs of analysis problems, and will give participants new insights regarding the types of issues that reviews can surface, the techniques to uncover them, and the methods to mitigate them (or where possible avoid them altogether) in one’s own work.
This tutorial requires no knowledge of any tools or languages. Only basic experience in statistical modelling and basic knowledge of statistics is assumed. However, to gain the most of the tutorial, participants will want to join with their brains firmly switched on, in order to reach, on their own, analytical insights as data is presented.
Bio: Dr Michael Brand is the Head and Founder of Otzma Analytics, a Data Science consultancy dedicated to maximising clients’ value from data by providing analytics upskilling, project review and executive mentoring. Before founding Otzma in 2018, Dr Brand served as Chief Data Scientist at Telstra, as Senior Principal Data Scientist at Pivotal, as Chief Scientist at Verint Systems, and as CTO Group Algorithm Leader at PrimeSense (where he worked on developing the XBox Kinect). Dr Brand also served as Director of the Monash Centre for Data Science in his role as Associate Professor for Data Science and AI at Monash University, where he remains an adjunct. Dr Brand holds a PhD in IT from Monash University, an MSc in Applied Mathematics from the Weizmann Institute of Science, and a BSc in Engineering from Tel-Aviv University. He has made industry-defining contributions that have earned him 18 patents (more pending), garnered many prestigious industry and academic awards, and power flagship products for the companies he worked with.