# Decisions in the Face of Uncertainty or Just Enough Statistics to be Dangerous

John Rauser (Snapchat)
Ballroom ABCD
Average rating:
(4.04, 47 ratings)

How many machines will we need for Q4? How big should our next datacenter build be? What are the odds this project will be completed on time? What does a network packet loss rate of 1% mean for my application?
Do you feel like you ought to have a working knowledge of statistics, but don’t know where to begin? If you took a statistics class in college, but have forgotten most of it, or you’ve never had any formal training in statistics, this class is for you. My goal is to give you a basic toolset to begin reasoning statistically through a series of interactive examples.

We’ll do an exercise in estimation, where you’ll learn how well calibrated you are as an estimator. We’ll use the data we get from the estimation exercise to explore one of the most useful ideas in statistics: the binomial distribution. We’ll conduct a key statistical test, the chi-square test, which use can to tell if the data you’re observing match what you’d expect or if the pattern is different. And last, we’ll solve a problem in decision theory and along the way learn about the normal distribution.

Though I’ll use a little bit of scary looking math, you should still attend if you’ve forgotten all the math you learned in school. I show how to use software like excel and sage to do all the mathematical heavy lifting.

## John Rauser

#### Snapchat

John has been extracting value from large datasets for over 20 years at hedge funds, small data-driven startups, Amazon, Pinterest, and now Snapchat. He has deep experience in machine learning, data visualization, on-line experimentation, website performance and real-time fault analysis. An empiricist at heart, “Just do the experiment!” is his favorite call to arms.

Suzanne Axtell
06/29/2011 3:31am PDT

Hi Nicolas, the video will be available to All Access and Online Access pass holders very soon.

Nicolas Poggi
06/28/2011 7:01pm PDT

Wow, this must have been a great presentation, will there be video? from the slides alone it hard to get all the info.

Aaron Peters
06/20/2011 1:48pm PDT

I liked John’s presentation, but I think people could not really get anything practical and useful from it. The formulas were overwhelming. I think John is an awesome speaker and he should be on stage every year, but this was a bit too much ‘theory’

Sophia DeMartini
06/20/2011 2:09am PDT

-Sophia

H. "Waldo" Grunenwald
06/19/2011 9:56pm PDT

I found this talk highly entertaining, and an excellent intro to Statistics. While I didn’t get as much from this talk as I would have liked, there were some key takeaways (such as that Average is seldom actually useful and to attempt to give a range rather than a single answer). Regardless, the entertainment value was high.

martin wignall
06/19/2011 7:06am PDT

Fantastic talk. Initially I wondered why I was in a statistics lecture, but actually (being an ex-mathematician anyway) realised that maybe I should dust off the old text books and apply this in my own work! Thank you for this wonderful exposition of a complex subject

sathya narayanan nagarajan
06/17/2011 9:01am PDT

It was great! I liked the way you presented!!!

John Rauser
06/17/2011 6:57am PDT

@Ernest: Good feedback. When I’ve given the talk to smaller audiences with a bit more time there’s more breathing room to work with the audience to address specific problems they have, but in 90 minutes and a giant room, that didn’t happen. Perhaps I’ll prepare another talk that more strongly favors application specific depth.

@Tim: I’ve given the presentation materials to the folks at O’Reilly; I’ll ask to have them posted.

James Schmidt
06/17/2011 6:37am PDT

John is an excellent presenter. This session gave me much to think about.

Tim Grant
06/17/2011 3:07am PDT

Will you be posting the slides and video of this session? I noticed that it was not in the Video and Slides section of the site… :(

Tim Grant
06/17/2011 3:06am PDT

John did an excellent job of clearly showing how to apply sophisticated statistical analyses to better diagnose complex situations. I wish my college statistics classes as clear cut… Thanks John

Ernest Mueller
06/15/2011 2:40am PDT

I enjoyed this session; my colleagues and I wished there had been more examples relevant to WebOps/Perf to drive it home how/when you might apply these principles on the job.

Steve Souders
05/19/2011 7:42am PDT

John was one of the highest rated speakers at Velocity 2010. He’s a data mining magician full time at Amazon. Metrics are critical for web performance – without them you’re flying blind. We all know some basics, but John will help everyone step up to the next level at this workshop. This is a must attend.