Tuesday, December 8, 2020

Final US Election Results - Polling Bias Update

I created the following video for my STAT 101 class, who get to learn about the topic of sampling bias. Polling bias (the difference between what the polls said would happen in the US election and what actually happened) is an example of sampling bias, so the lecture may be of interest to a much more general audience than just my students. You can check it out here:

Friday, November 6, 2020

First Post Election Update

 I created a second video on the US election for my class, which I thought might be an interesting follow-up to my final pre-election video, which I posted on youTube and published an article on it here (with an embedded video link).

To summarize, the election is not yet called, they are still counting votes. As of this morning, my election prediction of 269 electoral college votes for Biden and 269 for Trump (see video in above linked article) is remarkably close to how things appear to be panning out. My prediction that the election is very very close appears to have been correct, which was in sharp contrast to all mainstream pollsters, who were all predicting a major Biden victory. I did think Trump was going to scrape through the electoral college system and eventually win the presidency in a long drawn out process involving Congress, this appears to not be the case. Trump is trying to fight the results in court, so that may result in drawing out the process of having an official definitive winner. I mentioned in my prediction video that it is a tight race and either candidate was in a position to win. 

I have not updated the analysis to include polling bias for 2020 as the final vote counts are not yet in. There will be more updates in the future for my class, so I will likely add more posts here as well.

Here is the video from youTube:

Monday, November 2, 2020

US Election, Polling Bias, Calling the Election

I have been tasked with teaching Statistics 101 this term, which includes the subject of sampling bias. One of the most common examples of sampling bias is polling bias, which turned out to be very important in the 2016 US election. I selected this topic as an in-depth example for my class as the 2020 election will unfold as our course does.

Towards that end, I have been updating the class throughout the term as to changes in the polls as well as what the polls would be if we were to adjust them to compensate for the known per state polling bias observed in the 2016 election. 

I promised a final update just before the election, which is found below in an embedded video. The recorded video outlines the issues I have identified that may impact the upcoming election (repeat of previous polling bias, as well as Biden's regionally lopsided support that doesn't help him). The video discusses some of the potential impact of those issues, made some preliminary predictions, and then I even take a chance on calling a possible election outcome, even though, at this point in time, statistically speaking, either candidate is in a position to win the election, assuming 2016 polling bias repeats itself (see the video). (spoiler alert: the race is wide open and my observations these past 6 weeks have inclined me towards Trump barely winning, I also outline a highly plausible scenario where he does just that). 

If you are interested, see the following video for a detailed lecture on this topic that I prepared for my class. I thought other people might find it interesting, so I decided to post it (plus I've been neglecting this site for a long time). The topic should be accessible enough to a general audience, even though I'm addressing my class.

I will find it interesting to compare the various possible outcomes outlined in the video with the results of the election that will be reported tomorrow night and intend to follow up with more posts on this topic, regardless of the outcome.

If the embedding above doesn't work, you can follow this link to the youtube page where the video is embedded from: https://www.youtube.com/watch?v=iCRFnTNBm7c&t=141s