Coding a good data science model for work is not the same as programming a good data science challenge for an interview. This tutorial will cover what you need to know in order to make your challenge stand out from the crowd as well as highlighting common mistakes and pitfalls to avoid. We will be stressing best practices more than specific machine learning techniques. This tutorial assumes you have a working knowledge of Pandas and at least one plotting library.

During the tutorial you will be split up into groups to work through different parts of the data science takehome challenge. The code for the tutorial can be found at https://github.com/MichoelSnow/pydata_nyc_2019

The focus for this tutorial will be following topics:

Working with trick data

• Investigating a data set for natural and synthetic errors
• How to handle errors in the data

Outlining your Challenge

• What time requirements actually mean
• How much time you should spend on a challenge
• How to infer what you will be graded on
• How to properly allocate your time

Exploratory Data Analysis (EDA)

• What makes a good plot
• Choosing what to plot and how to plot it
• Knowing when you are done with EDA

Modeling data

• How to choose the appropriate model for the company
• How to explain model findings

If time permits we will also be discussing the following general best practices:

• How to make sure your code runs for your interviewer
• How to document and comment appropriately