Twitter | Search | |
Michael DeCrescenzo
PhD candidate (elections, Bayesian statistics, research transparency, voter ID). I like GIFs and graphs. alum.
1,714
Tweets
566
Following
564
Followers
Tweets
Michael DeCrescenzo 2h
I deleted some tweets because the threading looked weird but here is a graph instead. I'm writing about Bayesian causal inference in situations where you, the researcher, know more than your model does about the parameters you should find
Reply Retweet Like
Michael DeCrescenzo 14h
It's Friday and I'm not gonna tweet about Bayesian statistics even though I've got one locked and loaded
Reply Retweet Like
Michael DeCrescenzo Mar 22
Why limit ourselves to just horizontal space? Where can I find a dataset that contains the heights of residential dwellings in every county? "As we can see, the GOP candidate had a strong hold on the first and second floors, but barely competed anywhere else"
Reply Retweet Like
Michael DeCrescenzo Mar 22
Replying to @mikedecr
Reply Retweet Like
Michael DeCrescenzo Mar 22
Made it easier to tell what happened when I accidentally knit a file that I shouldn't knit
Reply Retweet Like
Michael DeCrescenzo Mar 21
can you believe how much better I feel because I was scolded into sitting up straighter
Reply Retweet Like
Michael DeCrescenzo retweeted
Dan Simpson Mar 19
We (, Andrew, Bob, and I) have a new paper out that talks about the convergence of MCMC. The quick summary is that R-hat (aka the Gelman-Rubin diagnostic) FAILS badly in some cases, so we fixed it. We also did a bunch of other stuff.
Reply Retweet Like
Michael DeCrescenzo retweeted
Hadley Wickham Mar 19
You may have heard a rumour that gather/spread are going away. This is simply not true (they’ll stay around forever) but I am working on better replacements which you can learn about at . Now is a great time for feedback!
Reply Retweet Like
Michael DeCrescenzo Mar 18
am I gonna regret hashtagging LaTeX because the internet is full of nightmares
Reply Retweet Like
Michael DeCrescenzo Mar 18
Obtaining entries from Google Scholar (& with "Google Scholar Button" extension []) is definitely easy, but the quality of existing entries is generally quite bad. Anybody have good (online) tools that combine ease with reliability?
Reply Retweet Like
Michael DeCrescenzo retweeted
WhyDoesR Feb 26
Reply Retweet Like
Michael DeCrescenzo Mar 17
Sublime's newest update has a block cursor
Reply Retweet Like
Michael DeCrescenzo Mar 16
This is intense. I am both amazed and terrified
Reply Retweet Like
Michael DeCrescenzo Mar 14
I think about this all the time. Some day the simple model will do you wrong and you won't know how to find out because you trained yourself not to care
Reply Retweet Like
Michael DeCrescenzo Mar 14
Peak laziness would be key-mapping `'` to `"` in Sublime because I want to print `"` but I only want to type `'`.
Reply Retweet Like
Michael DeCrescenzo Mar 12
will my students know what this is
Reply Retweet Like
Michael DeCrescenzo retweeted
Andrew Gelman Mar 12
R package for Type M and Type S errors
Reply Retweet Like
Michael DeCrescenzo Mar 10
A Bayesian framework has second-nature enhancements for causal mediation (in the Imai et al style) because you can estimate effects _marginal_ of the sensitivity parameter, not just conditional on it 🥳 All you need is priors about that param, which you proooobably already have!
Reply Retweet Like
Michael DeCrescenzo retweeted
Molly Beck Mar 9
Serious question: Will it ever stop snowing
Reply Retweet Like
Michael DeCrescenzo retweeted
Alice Evans Mar 9
Academics claim to be guided by evidence, yet venerate big names, are blinkered by bias, & perpetuate inequality. In this thread, self-interested editors seek Big Names (& are relaxed about intellectual theft?) Awesome allies ( & ) tell them to **** off. 👊
Reply Retweet Like