Andrew TherriaultOvjeren akaunt

@therriaultphd

Data scientist. Formerly led data science and analytics teams and . All opinions / corgi pix / bad puns are mine alone.

Boston, MA
Vrijeme pridruživanja: rujan 2010.

Tweetovi

Blokirali ste korisnika/cu @therriaultphd

Jeste li sigurni da želite vidjeti te tweetove? Time nećete deblokirati korisnika/cu @therriaultphd

  1. Prikvačeni tweet

    After wrapping up my Harvard course and one of my main consulting gigs in January, I'm looking to take on some new projects, and updated my site to show more of what I can do: . Please share, and reach out if you have ideas for collaborations. Thanks!

    Poništi
  2. The Xennial president we've been waiting for.

    Poništi
  3. Poništi
  4. Updated yet again, now that 96% of results are in. Same links as below. No big changes from earlier today, but a few. Probably won't post another update since the last 4% won't really change any conclusions, but may run it again tomorrow if there's demand - let me know?

    Poništi
  5. Poništi
  6. Prikaži ovu nit
    Poništi
  7. Ugh... of course they did: . I'll update if anything substantial changes, but hopefully it's just as minor as they say (18/17?!?)

    Prikaži ovu nit
    Poništi
  8. Oh FFS. I literally just published an updated analysis.

    Poništi
  9. So that's it for this edition of the madness that ensues when 150,000 unrepresentative people in a rural red state use an unnecessarily complex and undemocratic event to kick off the Democratic nominating process. Will update later if I find out more. Thanks for reading! (17/17)

    Prikaži ovu nit
    Poništi
  10. down the road if he's not the eventual nominee. Hopefully in that scenario it wouldn't be as bad as 2016, but still, it's something to be aware of as we watch these intra-party fights drag out for another month or two. And god help us if it's a contested convention. (16/x)

    Prikaži ovu nit
    Poništi
  11. the caucus system is problematic - if the goal of reallocation is to produce more consensus, it fails when there's an incentive to game the popular vote in this way. But if it's actually a sign of an unwillingness to compromise, that could cause a lot of headaches... (15/x)

    Prikaži ovu nit
    Poništi
  12. Given that these are caucus results, maybe this is some kind of strategic choice (e.g., not wanting to add to the popular vote totals of another candidate, which would hurt Sanders' overall proportion of the vote share). But even if so, that's yet another reason why... (14/x)

    Prikaži ovu nit
    Poništi
  13. at less than 10%. But the one exception to this? Sanders supporters. Based on my analysis, at least 23% of Sanders supporters chose *nobody* rather than switch to one of the other major candidates. So could we be seeing a repeat of 2016's "Bernie or Bust" movement? (13/x)

    Prikaži ovu nit
    Poništi
  14. If you add up those numbers, you get greater than 90% for 4 out of the 5 candidates' supporters switching, which sets the minimum estimate of drop-off (voters switching to Uncommitted or just going home rather than going to another major candidate) for those supporters... (12/x)

    Prikaži ovu nit
    Poništi
  15. total across other candidates to see where any candidate's supporters move to in general. But if you add up those numbers, you do get an upper bound for the percentage who switch over to *any* other of the major candidates, and that's where it gets really interesting. (11/x)

    Prikaži ovu nit
    Poništi
  16. Finally, on to the most interesting new thing that I alluded to earlier. The model-based analysis is a bit tricky to interpret in general terms, since it's based on specific scenarios (if x is not viable but y is, what % of x's voters does y get?). So you can't just... (10/x)

    Prikaži ovu nit
    Poništi
  17. to make their pitches to unallocateds *before* the first official count, which makes sense because you want to make sure you're viable first and unallocateds are the most logical persuasion targets. If so, then it makes sense that they'd pick up few of them later on. (9/x)

    Prikaži ovu nit
    Poništi
  18. Another pattern that I didn't think much about yesterday: why (according to the best estimates I can come up with) did Warren get so few of the initially unallocated, and Buttigieg appeared to get none? My guess is that those campaigns' ground games were good enough... (8/x)

    Prikaži ovu nit
    Poništi
  19. end up in the long run. (That's under the assumption that Yang doesn't somehow turn it around, but I'm not saying that's impossible! Don't really feel like getting ratio'd here today, thanks.) (7/x)

    Prikaži ovu nit
    Poništi
  20. Next: Looking at where Yang voters ended up, it's interesting that they seemed split between Sanders, Buttigieg, and Klobuchar, but didn't much like Biden or Warren. Those groups don't align with either the ideology or gender narratives, so it's unclear where those votes... (6/x)

    Prikaži ovu nit
    Poništi
  21. Right now, 538 has Klobuchar at about 9% in NH. If he managed to pull away a third of those voters over the next week, that could push him away from the Biden/Warren grouping and into direct competition with Sanders. So her future could directly affect his, and quickly. (5/x)

    Prikaži ovu nit
    Poništi

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·