Quantcast
Viewing all articles
Browse latest Browse all 44

A flaw in fivethirtyeight (and a meta-analytic repair)

This post continues my discussion of how to meta-analyze state polls. My most recent post was on Saturday morning. Today I’d like to focus on a particular strength of meta-analysis, its use as a precise tracking tool.

Meta-analysis of state-level polls allows one to make an estimate of where the race stands at any given moment. As a biophysicist and neuroscientist, I use simple yet powerful statistical tools that reduce the complexity of state polls to a single snapshot, the Median Electoral Vote Estimator. Its effective margin of error (MoE) is currently 35 electoral votes, equivalent to less than 1.0% in units of popular votes.

This exceptionally small MoE outperforms any other publically available measure, including fivethirtyeight.com. Poblano’s resource is valuable and fully deserving of its popularity. However, his work does contain a flaw that introduces unnecessary error - one that I fixed four years ago in 2004. My repair is available at election.princeton.edu for your use between now and November.


Viewing all articles
Browse latest Browse all 44

Trending Articles