20 Nov 2015
16:00 - 16:45
Chris Stucchio – Data science
What’s this all about data science? Explain baysian statistics to me as a kid – what should I know?
Scroll down to read the summary, watch the video and dowload the slides!
Summary Chris Stucchio
Bullet point summary by Theo van der Zee:
- Until about half a year ago, almost all A/B testing tools were using T-tests to determine winners.
- Frequentist statistics are often more difficult to explain that more modern statistical methods.
- Bayesian statistics: change your prior assumption based on new evidence, to a posterior situation.
- Conversion rates can be more accurately depicted as curves rather than discrete bars.
- As your gather more data, your posterior value is getting closer to true data value.
- Hypothesis testing was made for and by scientists, while CRO is more like trading.
- Identify a value of caring – a CR drop so small that you don’t care about it.
- Bayesian tests are insensitive to peeking – it’s fine to stop a test early.
- Chance to beat control is actually that, the chance to beat the control.