Hugo Bowne-Anderson interviewed me for a DataCamp podcast. Transcript is here.
3 thoughts on “Podcast interview on polling (mostly), also some Bayesian stuff”
A good read. One note: According to Nate Silver’s 2016 post-mortem, the adoption of odds (e.g. 2 in 5) over percentages for forecasting election outcomes seems to be more about avoiding a specific misperception than backing away from unrealistic precision.
“Also, both probabilities and polls are usually listed as percentages, so people can confuse one for the other — they might mistake a forecast showing Clinton with a 70 percent chance of winning as meaning she has a 70-30 polling lead over Trump, which would put her on her way to a historic, 40-point blowout. [in note:] For this reason, we may experiment with listing probabilities as odds — e.g., Trump has a 2 in 7 chance — rather than as percentages in future election years.” https://fivethirtyeight.com/features/the-media-has-a-probability-problem/
I remember seeing this when it was published because I’m interested in the usability of charts and quantitative measures. As of this writing, Democrats have a 20.8% chance of winning control of the Senate but the headline above that number says “1 in 5.”
2 in 7 is not odds, it’s probability expressed as a rational number, 2/7 the use of the word “in” clearly expresses the idea that 2 is the number of “successful possibilities” whereas 7 is the “total number of possibilities” so that 2 in 7 represents the ratio of successes to the total, or probability.
odds would be 2 to 5 against, there are 2 successes and 5 non-successes being considered, the probability is 2/(2+5) and the second number doesn’t express the totality of options (in 7) but rather the number of alternative outcomes (to 5)
A good read. One note: According to Nate Silver’s 2016 post-mortem, the adoption of odds (e.g. 2 in 5) over percentages for forecasting election outcomes seems to be more about avoiding a specific misperception than backing away from unrealistic precision.
“Also, both probabilities and polls are usually listed as percentages, so people can confuse one for the other — they might mistake a forecast showing Clinton with a 70 percent chance of winning as meaning she has a 70-30 polling lead over Trump, which would put her on her way to a historic, 40-point blowout. [in note:] For this reason, we may experiment with listing probabilities as odds — e.g., Trump has a 2 in 7 chance — rather than as percentages in future election years.” https://fivethirtyeight.com/features/the-media-has-a-probability-problem/
I remember seeing this when it was published because I’m interested in the usability of charts and quantitative measures. As of this writing, Democrats have a 20.8% chance of winning control of the Senate but the headline above that number says “1 in 5.”
2 in 7 is not odds, it’s probability expressed as a rational number, 2/7 the use of the word “in” clearly expresses the idea that 2 is the number of “successful possibilities” whereas 7 is the “total number of possibilities” so that 2 in 7 represents the ratio of successes to the total, or probability.
odds would be 2 to 5 against, there are 2 successes and 5 non-successes being considered, the probability is 2/(2+5) and the second number doesn’t express the totality of options (in 7) but rather the number of alternative outcomes (to 5)
“Transcript is here”
Found the audio but the Transcript was unavailable.