Skip to content

Instantly share code, notes, and snippets.

@rthadani
Forked from ezmiller/scicloj-workshop-template.md
Last active March 18, 2022 02:53
Show Gist options
  • Save rthadani/c4f51178b68962ac82114a8ab5d0e0cf to your computer and use it in GitHub Desktop.
Save rthadani/c4f51178b68962ac82114a8ab5d0e0cf to your computer and use it in GitHub Desktop.
Scicloj Workshop Template

Work In Progress

Scicloj Workshop: Introduction to the measures of statistical inference

Useful inferential statistics does not have to be just the domain of data scientists. This workshop follows examples in the book Statistics is Easy by Dennis Shasha and Manda Wilson, along with other real world examples to demonstrate concepts of fairness, p-value, confidence intervals, power. All concepts will be explained purely using functions from the core clojure library.

Few people remember statistics and often have to refer to books to apply formulae, hoping that the underlying distribution and its assumptions satisfy the needs of the formula, however the magic and the assumptions can be made concrete and clear by using randomization techniques collectively known as resampling.

In the workshop, we will simulate an experiment and analyze it statistically, answering the following questions: how accurate is the measurement likely to be (confidence interval), and could it have happened by mistake?

About the presenter

Rohit Thadani currently works at Barracuda Networks as a Prinicipal Engineer, working on platforms that collect store and make over 100 million emails searchable everyday.

Workshop Length & Structure

  • Two sessions 45 minutes each.
    • The first session will cover experiments used to understand significance tests most commonly used for AB experiments where an A is the normal occuring pattern and B is the observed pattern. We are trying to answer the question is B significantly different from A. We will talk about p-values in depth and look at some practical examples
    • The second session will include experiments that compare two groups to determine where each group has a range of values. In addition to saying wether the groups are significantly different from each other we will try to understand the measure of the difference between the groups(comfidence intervals) and concepts of false positives(power). If time permits we can talk about other practical considerations such as outliers.
  • Office hours. There will be a few simple exercises posted at the end of every session and if there is a need for discussion or setup related questions I can answer them. Also there are a bunch of interesting measures I wont have time to cover I can mention them here.

Where?

Recordings

Most but not all of the workshop session(s) will be recorded. We will announce when the recording stops & starts.

Signup

To sign up for the workshop: Click the Going button at the top of this post. (You can also use the Add to Calendar to add the event to your calendar.)

We will share a link with participants in the thread 48 hours before the workshop. We recommend following the #scicloj-workshops thread on Zulip. The link will also be shared there.

@ezmiller
Copy link

Looks great @rthadani !

A few minor suggestions/alterations to consider:

Paragraph 1:

  • "This workshop will follow examples... "

Paragraph 2:

  • "...hoping that the underlying distribution and its assumptions satisfy the needs of the formula. However, the magic and the assumptions..."
  • "...the magic and assumptions can be made concrete and clear by using a randomization technique..."

Paragraph 3:

  • "In the workshop, we will simulate an experiment and analyze it statistically, answering the following questions: how accurate is the measurement likely to be (confidence interval), and could it have happened by mistake?"

@ezmiller
Copy link

@rthadani, a few other thoughts

  • Would it make sense to specify what the two 45-minute sessions will be about?
  • I think you should explain a bit more what "office hours" means. What are the office hours for?
  • It think it would make sense to link the books you mention.

Anyway, this looks close. Seems like you'll be ready to post this on Sunday. @elenam have you taken a look?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment