Tweet
As a baseball fan, I'm a longtime fan of Nate Silver. His contributions to a deeper understanding of a highly random, nuanced game through insightful statistical modeling are undeniable and admirable.
When he ventured into political poll meta-analysis and predictions at his website, www.fivethirtyeight.com, it put a similarly new spin on politics. He has a pretty strong track record in this area overall (ask any Romney fan). But this year's ongoing analysis has been full of odd results, and Nate's reactions in interviews have been a bit on the testy side.
Following is an analysis of Nate's analytics by a UIC statistics prof. Yes, it's long, and yes, it's from the dreaded Huffington Post. But it's a great read for stats nerds. Maybe greater after the election results are done.
Posted as a public service..
[massively snipped]
TL/DR summary: 538's constantly-updated presidential election meta-poll analysis is pretty opaque. The author posits the likely flaws in Nate Silver's statistical modeling methodology that best explain why 538 has consistently shown a higher probability of a Trump win than any of the other meta-polling sites out there, and has also exhibited some notably odd jumps that seem to be at odds with news cycles and other polls. Basically, it looks like Silver has an opaque, overly complex model that has too many moving parts and tends to squeeze to the middle.
Full article:
I found it to be a refreshing (nearly) apolitical story, perfect for capping off the national nightmare that has been the U.S. presidential election season of 2016. Very weird to see such a thing on HuffPo.
When he ventured into political poll meta-analysis and predictions at his website, www.fivethirtyeight.com, it put a similarly new spin on politics. He has a pretty strong track record in this area overall (ask any Romney fan). But this year's ongoing analysis has been full of odd results, and Nate's reactions in interviews have been a bit on the testy side.
Following is an analysis of Nate's analytics by a UIC statistics prof. Yes, it's long, and yes, it's from the dreaded Huffington Post. But it's a great read for stats nerds. Maybe greater after the election results are done.
Posted as a public service..
I’m A Stats Prof. Here’s Why Nate Silver’s Model Was All Over The Place.
538’s presidential model is built in a sub-optimal way. It’s at once far too complex, while also making basic errors that have thrown its projections out of whack.
11/08/2016 04:02 pm ET
By: Dale Rosenthal
Clinical Assistant Professor of Finance, University of Illinois at Chicago
Now that the votes are finally being cast and counted, it’s time to turn to that hallowed American tradition: rating how well pollsters and meta-pollsters did. I cannot claim to have an opinion on how most of these performed. However, I did take a look into one of the best-known meta-pollsters: Nate Silver’s FiveThirtyEight.com. Unfortunately for 538, it seems we can call the race early: things did not go well.
Silver’s model, throughout the campaign, showed a dramatically higher likelihood that Donald J. Trump would be our next president ― higher than any other of the major forecasting models at work. That might be OK (and I’ll get to why shortly), but his model also sometimes exhibited strange behavior ― like moving overall results the opposite way of new data fed into the model. Silver has defended his projection as “commonsense” and dismissed models that estimated probabilities above 90% that Hillary Clinton would win.
That strange behavior is what initially led me to take a closer look at Silver’s methodology.
538’s presidential model is built in a sub-optimal way. It’s at once far too complex, while also making basic errors that have thrown its projections out of whack.
11/08/2016 04:02 pm ET
By: Dale Rosenthal
Clinical Assistant Professor of Finance, University of Illinois at Chicago
Now that the votes are finally being cast and counted, it’s time to turn to that hallowed American tradition: rating how well pollsters and meta-pollsters did. I cannot claim to have an opinion on how most of these performed. However, I did take a look into one of the best-known meta-pollsters: Nate Silver’s FiveThirtyEight.com. Unfortunately for 538, it seems we can call the race early: things did not go well.
Silver’s model, throughout the campaign, showed a dramatically higher likelihood that Donald J. Trump would be our next president ― higher than any other of the major forecasting models at work. That might be OK (and I’ll get to why shortly), but his model also sometimes exhibited strange behavior ― like moving overall results the opposite way of new data fed into the model. Silver has defended his projection as “commonsense” and dismissed models that estimated probabilities above 90% that Hillary Clinton would win.
That strange behavior is what initially led me to take a closer look at Silver’s methodology.
TL/DR summary: 538's constantly-updated presidential election meta-poll analysis is pretty opaque. The author posits the likely flaws in Nate Silver's statistical modeling methodology that best explain why 538 has consistently shown a higher probability of a Trump win than any of the other meta-polling sites out there, and has also exhibited some notably odd jumps that seem to be at odds with news cycles and other polls. Basically, it looks like Silver has an opaque, overly complex model that has too many moving parts and tends to squeeze to the middle.
Full article:
I found it to be a refreshing (nearly) apolitical story, perfect for capping off the national nightmare that has been the U.S. presidential election season of 2016. Very weird to see such a thing on HuffPo.
Comment