Twitter | Search | |
Mike DeCrescenzo
Political science PhD candidate, UW–Madison. Researches elections, voter ID, statistics of a Bayesian flavor. Also: bikes, , GIFs
1,400
Tweets
534
Following
497
Followers
Tweets
Mike DeCrescenzo Aug 17
Replying to @BenListyg
We do a few days of practice with pre-calc and calc things, set theory, linear algebra basics, and probability basics. Can't get deep into actual stats but we do a few Bayes theorem examples
Reply Retweet Like
Mike DeCrescenzo Aug 17
Replying to @mikedecr
You know I had to do it to 'em
Reply Retweet Like
Mike DeCrescenzo Aug 17
Dear diary, Day z of the resistance against the frequentist menace. A new battle looms. PhD program math bootcamp begins Monday.
Reply Retweet Like
Mike DeCrescenzo Aug 17
Reply Retweet Like
Mike DeCrescenzo Aug 17
exactly the kind of notes from past-self that present-self loves to find
Reply Retweet Like
Mike DeCrescenzo Aug 17
Replying to @chagai_weiss
My default approach is that packages are handy for data processing but I don’t really prefer to use a graphic that a package made for me. I’d prob use broom::tidy() for coefs/intervals and broom::augment for predictions, which return data that are easily fit into ggplot
Reply Retweet Like
Mike DeCrescenzo Aug 16
hmmm, is that so
Reply Retweet Like
Mike DeCrescenzo Aug 16
Thanks! I actually had a lot of trouble w/ this file since it's v different from what the toolchain sets up for me ordinarily. I currently have mine set up w/out the additional recommendations to modify CXXFLAGS and so on and yet I'm finding the brms model to be faster rn
Reply Retweet Like
Mike DeCrescenzo Aug 15
the models that work are much faster though so ¯\_(ツ)_/¯
Reply Retweet Like
Mike DeCrescenzo Aug 15
mm ya shouldn't have done that. walked face first into the "unknown C++ exception" hole again even when running things that used to work
Reply Retweet Like
Mike DeCrescenzo Aug 15
Reply Retweet Like
Mike DeCrescenzo Aug 15
This is making me seriously paranoid that I've been running Stan with a bad under-the-hood setup for a long time. I can't even find a makevars file to investigate this 😱
Reply Retweet Like
Mike DeCrescenzo Aug 15
thinking about how you would find the probability that two people play a game of chutes and ladders that never ends
Reply Retweet Like
Mike DeCrescenzo retweeted
Aki Vehtari Aug 6
Bayesplot 1.6.0 in CRAN - no longer overrides the ggplot theme fix by Malcolm Barrett - improved visual diagnostic vignette by - more arguments to various functions See more details in NEWS
Reply Retweet Like
Mike DeCrescenzo Aug 14
Replying to @tjmahr
If you want to avoid reserved terms then isn’t “consistency” also a problem? I might go with “likely” because it kinda means the same thing even in itsnreserved sense
Reply Retweet Like
Mike DeCrescenzo Aug 14
While I would never claim that it’s _impossible_ for a politician’s favorite Wisconsin beer to be Spotted Cow, i will claim that if you’re a politician and you say that your favorite Wisconsin beer is Spotted Cow, you are probably lying
Reply Retweet Like
Mike DeCrescenzo Aug 9
Best to avoid rushing to conclusions. elections are big administrative undertakings and small errors with non-malicious causes are not rare
Reply Retweet Like
Mike DeCrescenzo Aug 9
While it is waaaay too soon to say for sure what the cause of this is, it's worth pointing out the irony with respect to Kobach's interest in strict photo ID: fraud (or error) is much cheaper and way more consequential when votes are being *counted*, not when they're being cast
Reply Retweet Like
Mike DeCrescenzo Aug 9
Difference between the uncertainty bands is striking though. Kinda makes you worry about how routinely we rely on the quick-and-dirty loess line in other situations!
Reply Retweet Like
Mike DeCrescenzo Aug 9
Narrator: it did not work the first time
Reply Retweet Like