tag:blogger.com,1999:blog-6976144462093297473.post7818098020830420864..comments2024-03-26T19:10:00.791-04:00Comments on West Coast Stat Views (on Observational Epidemiology and more): Evaluating evidenceJosephhttp://www.blogger.com/profile/10760453165301871031noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-6976144462093297473.post-6948859274311088202011-11-24T10:49:54.385-05:002011-11-24T10:49:54.385-05:00It would be nice if Mike Kimmel's work was to ...It would be nice if Mike Kimmel's work was to make it into the national debate. The idea that we are willing to SACRIFICE economic growth to have a low marginal tax rate on high earners would certainly change the color of the debate.Josephhttps://www.blogger.com/profile/10760453165301871031noreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-44140842829406123872011-11-24T00:19:56.478-05:002011-11-24T00:19:56.478-05:00Mike Kimmel at Angry Bear has done the obvious reg...Mike Kimmel at Angry Bear has done the obvious regressions using 20th century data, and no matter how he slices it, peak marginal rates around 60-70% maximize economic growth. He's tried to eliminate confounding factors, and I imagine his confidence level is low, but sometimes you have to work with the data you have. People don't drink gasoline based on less data than that.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-79391589078257570422011-11-20T21:04:07.415-05:002011-11-20T21:04:07.415-05:00One thing that comes up in election forecasting is...One thing that comes up in election forecasting is: What level of confidence do you want with for your national forecasts? I don't think 95% intervals make sense. Why? Because 20 elections will take 80 years. In 80 years you'd expect a (statistical) regime change of some sort. So I think it doesn't make much sense to ask for that level of confidence in a mechanical forecast.Andrew Gelmanhttps://www.blogger.com/profile/02715992780769751789noreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-65640827088680575292011-11-20T11:34:00.902-05:002011-11-20T11:34:00.902-05:00Agreed. I think that these inference approaches a...Agreed. I think that these inference approaches are especially important in data with high importance but low numbers. For example, I wonder if presidential election forecasting would not run into the same problems?Josephhttps://www.blogger.com/profile/10760453165301871031noreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-10146402311349812952011-11-20T11:31:10.049-05:002011-11-20T11:31:10.049-05:00It's a common-sense principle but I've nev...It's a common-sense principle but I've never seen it formalized or put in any book.Andrew Gelmanhttps://www.blogger.com/profile/02715992780769751789noreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-84151192450166307122011-11-20T03:26:43.486-05:002011-11-20T03:26:43.486-05:00Andrew,
What's discouraging is how seldom thi...Andrew,<br /><br />What's discouraging is how seldom this common sense principle is applied to the public discourse.Markhttps://www.blogger.com/profile/14705408455380402571noreply@blogger.comtag:blogger.com,1999:blog-6976144462093297473.post-50364284603800123792011-11-19T20:23:02.372-05:002011-11-19T20:23:02.372-05:00This reminds me of a point I often make to grad st...This reminds me of a point I often make to grad students: if you have a sample that is too small to confirm a hypothesis or fit a model, it can still be enough to allow you to reject some simple hypotheses which, in the absence of that small dataset, might otherwise seem potentially plausible.Andrew Gelmanhttps://www.blogger.com/profile/02715992780769751789noreply@blogger.com