![Andrew Gelman](https://i1.feedspot.com/57686.jpg?t=1659933450)
Andrew Gelman
3,334 FOLLOWERS
Blog on Statistics by Andrew Gelman, a professor of statistics and political science and director of the Applied Statistics Center at Columbia University. He has received the Outstanding Statistical Application award from the American Statistical Association.
Andrew Gelman
15h ago
The Republican vice-presidential nominee made some remarks a few years ago about non-parents (“We are effectively run in this country . . . by a bunch of childless cat ladies who are miserable at their own lives and the choices ..read more
Andrew Gelman
1d ago
There are so many interesting and important things to do in statistical modeling, causal inference, and social science, and so many places for recent graduates to jump in. Here are two opportunities that happen to have come in the mail ..read more
Andrew Gelman
3d ago
Recently in the sister blog:
A hallmark of human cognition is the capacity to think about observable experience in ways that are nonobvious—from scientific concepts (genes, molecules) to everyday understandings (germs, soul). Where does this capacity come from, and how does it develop?
Contrary to what is classically assumed, young children often extend beyond the tangible “here-and-now” to think about hidden, invisible, abstract, or nonpresent entities. . . . The standard developmental story may be backward: for young humans, going beyond the obvious can be easy, and sticking with the here-a ..read more
Andrew Gelman
4d ago
If you’re not careful, you’ll spend most of your time on computations that don’t work.
Here’s how we put it in the Bayesian workflow article:
An important intermediate goal is to be able to fail fast when fitting bad models. This can be considered as a shortcut that avoids spending a lot of time for (near) perfect inference for a bad model. There is a large literature on approximate algorithms to fit the desired model fast, but little on algorithms designed to waste as little time as possible on the models that we will ultimately abandon. We believe it is important to evaluate methods on thi ..read more
Andrew Gelman
4d ago
I have a feeling that our regular readers are getting tired of these election-news posts, so I’ll pop it in the middle of the night U.S. time so it will be easy to skip. I hope you’ll be interested in our post that will appear later this morning, “The ‘fail fast’ principle in statistical computing.” If you want to know what posts are coming a week or two in advance, just subscribe to our newsletter.
If you care enough about politics to have continued to read this far, you’ll be disappointed in the content of this post. Not because I’m gonna tell you that Harris will be a great candidate, and ..read more
Andrew Gelman
5d ago
A couple days ago, Raphael asked in the comment thread:
I have a question about a hypothetical scenario. If one of the candidates were to leave the ticket and be replaced by another candidate, would you recalibrate the model to match the new candidate? I mean, throw out the old Biden vs. Trump polls and wait for new ones? Or would you say, along the lines of your argument about 538’s high uncertainty, that the model allowed for the possibility of a candidate switch? Or, third option, try both (old calibration and new calibration) to see if the model output changes in any meaningful way at all ..read more
Andrew Gelman
6d ago
Nathan Yau puts it well:
No single chart type can show every angle of every dataset all the time. Every chart type has its trade-offs. So instead of trying to show everything at once, use multiple views to show things separate.
As I said in an earlier discussion, one thing that bothers me about the famous Napoleon-in-Russia graph is that it’s led people to suppose that it’s generally a good idea to convey a complex multidimensional story in a single plot. I think lots of data stories are just better told in multiple graphs. Someone could be the greatest designer in the world but it doesn’t m ..read more
Andrew Gelman
1w ago
I just read some books. Rather than review or summarize them, I’ll just pick out one or two things from each:
The Good Word, by Wilfrid Sheed (1978): Mostly book reviews. Here’s a good line: “All available evidence suggests that everyone was trying to be witty in the twenties, just as everyone was trying to be authentic in the sixties. Each era is bullied by one temperament which seems like the only one to have.”
One thing I enjoyed about this book is how relaxed Sheed seems to be. He’s not climbing the greasy pole of success, he’s not worried about losing his job or his cultural cachet, he’s ..read more
Andrew Gelman
1w ago
Someone who would like to remain anonymous writes:
I have worked on a number of equity agreements as part of the collective bargaining process at universities and feel that the methodology being used in that setting is confused at best.
Typically, salaries are regressed against a variety of variables, including gender. In the simplest kind of study, the coefficient of the gender variable is called the gender penalty (or bonus). I have seen at least one peer reviewed study that claimed that because the regression was done on the entire faculty body, there can be no concept of error of the coef ..read more
Andrew Gelman
1w ago
OK, this is all getting a bit recursive, but I think there are some more insights to be squeezed out of this particular lemon, so here we go . . .
Yesterday I wrote a long post on polling averages and political forecasts, focusing on the differences between Fivethirtyeight, whose model gives each party a 50% chance of winning, and the Economist, where the Republicans are at 75%.
In comments, Michael Weissman pointed to a blog post yesterday by Nate entitled, “Why I don’t buy 538’s new election model: It barely pays attention to the polls. And its results just don’t make a lot of sense.”
I’m gl ..read more