Thread by @mjskay

Really excited to announce our (Pierre Dragicevic, Yvonne Jansen, @abhsarma, me, @zirtech) paper on Explorable Multiverse Analyses. We make multiverse analyses (a la Steegen, @StatModeling, et al) interactive through explorable explanations (a la @worrydream). THREAD ---> 1/7

There is a multiverse of possible ways data in a paper could have been transformed, analysed, or reported. Instead of a stale old PDF, what if you could change these options to inspect that multiverse? 2/7

As an author, you might wish to provide a variety of analyses you consider reasonable. As a reader, you might wish to judge how robust an analysis is to these alternatives. Animation can help explore the multiverse --- this particular interaction may not be so robust. 3/7

In the spirit of specification curve (@uri_sohn, @jpsimmon, Leif Nelson), what if you could see an overview of several possible analyses, then click on each to drill down and see details? 4/7

In a Bayesian analysis, what if the reader could apply their own prior to a focal parameter, like a mean difference between two treatments? (R package by @abhsarma for this coming soon --- you'll be able to do it interactively in @rstudio and make notebooks with this widget!) 5/7

What if your report does not easily admit uncertainty visualization --- like the plot of conditions ranked by performance in this paper. Maybe it would be helpful to bootstrap your entire paper? (Hypothetical outcome papers! cc @JessicaHullman) 6/7

Check out the complete demos here: explorablemultiverse.github.io

And read more details and discussion (like just how should you write a conclusion section for a paper like this?) in our paper: hal.inria.fr/hal-01976951/d… 7/7

Forgot to mention! We are honored that this paper is getting a Best Paper award at #chi2019 (yay!). Now that the #chi2019 #templatewars have concluded, isn't it time to think about what interactive papers should look like? And shouldn't we be the community to do it? @sig_chi