David Robinson

We ran into this problem often teaching R when I was in grad school. However, in the years since, the tidyverse has put strong focus on getting users manipulating data even if they don't know how to define a function, etc..! Here's a useful post, comparing the classic approach you mention to an alternative http://varianceexplained.org/r/teach-tidyverse/

I created an open source A/B test framework[0], which also uses Bayesian analysis on the dashboard. IANAS(tatistician), but from what I understand it’s still better to plan the check point in advance, rather than stop when reaching significance. A couple of articles worth reading [1] [2] (can’t exactly vouch for their validity but seem to make some good arguments that appear thought out) [0] https://github.com/Alephbet/gimel [1] http://varianceexplained.org/r/bayesian-ab-testing/ [2] http://blog.analytics-toolkit.com/2017/the-bane-of-ab-testing-reaching…

I was thinking of the same article, and found it, I think [1]? There was a higher-rated link also posted around the same time with the same post content [2]. I also found a follow up post [3] while looking for [1]. [1] http://varianceexplained.org/r/trump-tweets/ [2] https://www.r-bloggers.com/text-analysis-of-trumps-tweets-confirms-he-… [3] http://varianceexplained.org/r/trump-followup/

Although tangentially linked to in the article, David Robinson's Introduction to Empirical Bayes[1] is also an excellent resource. It deals primarily with beta-binomial distributions. [1] http://varianceexplained.org/r/empirical-bayes-book/

I'm just going to wait it out, as these trends come and go; before AI it was Rust, before Rust it was IoT, before IoT it was big data, crypto was weaved through it, etc. I'm sure someone's done the numbers on HN trending topics over time aaand yup http://varianceexplained.org/r/hn-trends//.

> From a Bayesian perspective the answer is “early peeking is fine”. That’s definitely not true. Early peeking increases your type I error rate, even with Bayesian methods. For a full explanation see http://varianceexplained.org/r/bayesian-ab-testing/

I don’t have the means (or time or interest) to do this analysis myself, but it has been claimed that half that accounts’ tweets are him personally and the other half are a team: http://varianceexplained.org/r/trump-tweets/

Disclaimer: arguably a quant but definitely not this kind. In this specific case you only need access to Trump his tweets, which you can get without parsing and evaluating the firehose (expensive to get and process). > In particular, basic NLP and sentiment analysis is pretty easy, especially like in this case when you're dealing with a very limited vocabulary. Exactly. With Trump's tweets you can probably find easy correlations between sentiment (good or bad) and a select range of words as well as tweet times, whether he is in office or not, whether it was tweeted from his android[1] or not. Statistically speaking you can also look at his overall sentiment. Eg. simply always short the market whenever he tweets might be the profitable move 80% of the time. [1]: http://varianceexplained.org/r/trump-tweets/

A related post is http://varianceexplained.org/r/teach-tidyverse/

Sometimes a good analogy can take the place of a visualization [1], albeit I did imagine a distribution getting more concentrated. [1] http://varianceexplained.org/statistics/beta_distribution_and_baseball…

Here is the link: https://www.taggernews.com/ and here is the article describing how it was built: http://varianceexplained.org/programming/tagger-news/

I thought it was determined (well, alleged, with data) that Trump uses an Android device[1]. Has that changed? [1]: http://varianceexplained.org/r/trump-tweets/

I found this a nice way to communicate the difference between AI, ML and Data Science [0]. TL;DR AI produces actions ML produces predictions Data Science produces insights [0] http://varianceexplained.org/r/ds-ml-ai/

Direct link to a scatter plot (x axis: quantity, y axis: gender): http://varianceexplained.org/figs/2017-04-27-tidytext-gender-plots/tot…

Just to expand on this, David Robinson quantified this a while back: http://varianceexplained.org/r/bayesian-ab-testing/ He ran some simulated A/B tests using my Bayesian A/B testing technique (which is now powering VWO). He showed that while peeking does ensure that loss < specified threshold, if you don't peek the loss will be even lower. So although peeking is still valid, that validity does come at a cost.

That would go against most public accounts of how Trump has been using his phone for tweeting. For one thing, Obama's personal account (i.e. not @POTUS, but @BarackObama) explicitly says in its profile: "This account is run by Organizing for Action staff. Tweets from the President are signed -bo. [0] No such disclaimer has been made about Trump's account, which tweets with the same characteristics and tone as it did well before his current presidential campaign [1]. You're right, though, that Trump's team does tweet for him on occasion. Either that, or Trump, who is famously not a computer user, has mastered the art of using two different mobile operating systems [2] [0] http://imgur.com/a/ZGyzo [1] https://github.com/sashaperigo/Trump-Tweets [2] http://varianceexplained.org/r/trump-tweets/

He probably does use his phone to tweet http://varianceexplained.org/r/trump-tweets/

Can we link to the original blog post please? http://varianceexplained.org/r/trump-tweets/ Not only is it nicer to the author, it's far better formatted as well.

Took me a while to find the Trump Tweets blog you referred to. It is here for anyone trying to find it: http://varianceexplained.org/r/trump-tweets By the author of the tutorial, not the poster of the link.

> For that, I might look at another source of data, Stack Overflow Careers profiles, and see which technologies tend to be used by the same developers > http://varianceexplained.org/images/network2.jpeg This shows "git" and "github" in a separate cluster from "C++" and "Python"? I don't understand this. These tools are used regardless of what other technologies are being used. For example, there are many Python and many C++ projects on github.

Thank you for explanation. If I understand correctly, you use this approach to increase sensitivity (compared to NHST) using the same data. Most probably, in your case, higher sensitivity (or power) comes at the cost of higher type I error rate. And this might be fine. Sometimes making more changes and faster is more important than false positives. In this case, you can just use a higher p-value threshold in the NHST framework. You might argue that the discrete type I error does not concern you. And that the potential loss in metric value is what matters. This might be true in your setting. But in real life scenarios, in most cases, there are additional costs that are not taken into account in the proposed solution: increased complexity, more time spent on development, implementation, and maintenance. I suggest reading this old post by David Robinson: https://varianceexplained.org/r/bayesian-ab-testing/ While the approach might fit in your setting, I don't believe most of other users of tea-tasting would benefit from it. For the moment, I must decline your kind contribution. But you still can use tea-tasting and perform the calculations described in the whitepaper. See the guide on how to define a custom metric with a statistical test of your choice: https://tea-tasting.e10v.me/custom-metrics/