Twitter | Search | |
Michael DeCrescenzo
PhD candidate, political science, UW–Madison. I study elections, voter ID, statistics of a Bayesian flavor. Also: bikes, , GIFs
1,489
Tweets
543
Following
504
Followers
Tweets
Michael DeCrescenzo 7h
Replying to @andrewheiss
mmm I can put it in a gist tomorrow morning. I should get back to work!
Reply Retweet Like
Michael DeCrescenzo 8h
Replying to @andrewheiss
Ok fixed. In this fig, y is (beta * humidity), plus the regression residual. This is equivalent to starting with the fully estimated regression and subtracting out terms for every other covariate
Reply Retweet Like
Michael DeCrescenzo 8h
Replying to @andrewheiss
Variation in y as f(x) but the level of y is wrong I think
Reply Retweet Like
Michael DeCrescenzo 8h
Replying to @andrewheiss
Here's how this looks when you isolate the effect of Humidity above & beyond the other terms in each model. The math might not be 100 right...but wanted to see what this looks like when the data are purged of the other dimensions
Reply Retweet Like
Michael DeCrescenzo 14h
setting aside everybody's traumatic memories of the needle, "the probability that [x] wins, given the predictiveness of the reported precincts" makes sense
Reply Retweet Like
Michael DeCrescenzo Oct 17
dreaming of a time when more journals will just accept an .md document and build it with their own template file rather than asking authors to do all the work to place tables and figs in separate files with specified dpi and color models
Reply Retweet Like
Michael DeCrescenzo Oct 17
Replying to @andrewheiss
Dear editor, do you accept appendix in microfiche?
Reply Retweet Like
Michael DeCrescenzo Oct 17
Replying to @JoshDeutschmann
IMO the biggest draw for me is being able to reference results using R objects directly in the text. I was exporting things to .tex files before but this is way smoother. Main downside is that there are more potential failure points during compilation, also startup costs
Reply Retweet Like
Michael DeCrescenzo Oct 16
Replying to @JoshDeutschmann
Check out "The Plain Person’s Guide to Plain Text Social Science"
Reply Retweet Like
Michael DeCrescenzo Oct 16
Replying to @JoshDeutschmann
sensible sensible... on the broader pt, you always wonder if the technical tool is justified or if it's overkill. But realistically..you do technical work, you have technical needs, and pop tools will fail in a bind if they aren't designed for you (which they aren't)
Reply Retweet Like
Michael DeCrescenzo Oct 16
Replying to @JoshDeutschmann
You use Atom IIRC? Their Git integration is awesome. I use GitKraken, I think it's a sensible interface that lets you view diffs and edit somewhat in a sensible way
Reply Retweet Like
Michael DeCrescenzo retweeted
Elections Research Center Oct 16
Register for our post-election symposium on December 7th ! Experts will analyze the results of the 2018 midterm elections and discuss a range of topics. Check out the planned panels and register for the free event at
Reply Retweet Like
Michael DeCrescenzo Oct 16
Insights from 's 2016 forecast about how not-crazy the 2016 outcome was. Given: • 10% (marginal) probability that Clinton loses electoral college but wins popular vote • 30% prob Trump wins EC ⇒ if Trump wins EC, 1-in-3 chance he wins by losing popular vote
Reply Retweet Like
Michael DeCrescenzo Oct 15
Replying to @jlenriquezch
Does this get too meta if I were to allege that the definition of p(x) is incorrect
Reply Retweet Like
Michael DeCrescenzo Oct 15
Awesome!
Reply Retweet Like
Michael DeCrescenzo retweeted
Steve Reilly Oct 12
Replying to @mikedecr @abtran
Highly recommend the lessons and readings from 's
Reply Retweet Like
Michael DeCrescenzo Oct 12
Replying to @kearneymw
I remember initially being concerned that R4DS assumes more comfort w/ the computer and w/ statistics than my students would have. Do you have experience teaching from this book? Are these hesitations off base?
Reply Retweet Like
Michael DeCrescenzo Oct 12
Authors from diverse and underrepresented backgrounds greatly encouraged! I'm also all-ears for any pedagogical advice about teaching quant things (consuming and creating quant content) to undergraduates!
Reply Retweet Like
Michael DeCrescenzo Oct 12
I may teach a "politics & data" course soon. I'd love to hear your recs for: - Intro R texts (cheap or free, prefer tidyverse/ggplot2, practicality over CS-heavy theory) - undergrad-accessible readings to show how taking data seriously yields neat and/or unexpected insights
Reply Retweet Like
Michael DeCrescenzo Oct 12
A fun thing you could do right now would be to go outside and ask any person what their profession is and if they think a randomly selected person would be able to do it better
Reply Retweet Like