I told you so.
In case you missed it, Sam Wang at Princeton Election Consortium and Nate Silver of FiveThirtyEight among others correctly predicted the outcome of Tuesday’s presidential election. And not only “correctly predicted” but predicted to a highly accurate degree the final electoral counts (and how FLA was a coin toss, statistically) and the popular votes.
In case you’re wondering, those screams you’ve heard around Charlotte, NC for the last two weeks have been me every time I’ve heard or read some media report going into the election that it was a close race, a toss up, and we had no idea what was going to happen. When the statistical analysts are coming back with 90% (Nate Silver) to 99 to 100% (Sam Wang) chance of the president winning, we have a pretty freaking good idea of what the data are saying.
Let’s put it this way, the next time, it calls for a 99% chance of rain, and you don’t carry an umbrella because you really, really want for it to be sunny? I will have a similar reaction.
So what are the takeaways?
1) Feelings are not facts. I got this from a friend who is a therapist when we were discussing what the pundits were predicting versus what the data were predicting. Somewhere along the way, American society has equated opinions with data. They are not the same. And in cases, of oh, I don’t know, smoking and health outcomes, climate change, or evolution, you can hold on to your deeply held beliefs, but if they don’t match the preponderance of the data, your beliefs are wrong.
2) Combined data is better than single data sources. Any one data source can have problems. Indeed, Gallup was off. WAY off. Why? Because they significantly overestimated the white turnout. That does not mean one should ignore data that doesn’t agree with your opinions or even the rest of the data. The cool thing about data aggregation is that it includes all the data and lets the errors/assumptions/sampling quirks statistically cancel each other out.
3) Learn how to trust and doubt at the same time. This is an advanced smarty pants move, and something I want to credit the writings and theorizing of Karl Weick. I also credit Public Image Limited for the same sentiment, but Weick is cited more academically. What it means is that you should believe your beliefs and be open to them being wrong. You should use your data, but it may have errors that haven’t been accounted for. You should, essentially, not believe much as being 100% true and should be open to, if not looking for, information to adjust your beliefs and improve your data. And if you don’t? If you are someone who only wants to hear about data that supports your beliefs and purposefully ignore the rest the data? You scare the crap out of me. Or, actually, it’s makes it harder to gain the respect of others.
So, there you go. Data, once again, remain neutral. Science/statistics/math can help us uncover the truth. We should all try to find some data that don’t support our beliefs and cogitate on them for a while.