November 4, 2020
The Evidence Workshop hails from the great state of Wisconsin where the results of the 2020 presidential election are too close to call. They say it’s a nail biter. We’re tempted to scroll Fox News and CNN all day to try to predict which way our state will go. Instead, we decided to use the close race to talk about sampling theory and generalizability.
Samples & Populations
When researchers want to study something, it’s almost always impossible to examine the entire population. For example, we cannot access all the people with Parkinson’s disease. So, we examine a subset of the population, called a sample. An ideal sample will represent the condition in its entirety – young and old, severe and mild, etc. Most samples are not ideal. They do not represent the population in its entirety. Therefore, we cannot generalize the results to the entire population. For example, if we sample only young, mildly impaired people with Parkinson’s disease we can only extend our results to young, mildly impaired people with Parkinson’s disease.
What does this have to do with the election?
One of the reasons that Wisconsin is too close to call is that voters in our state are split approximately 50/50 in their preference for the two presidential candidates. Experts argue that “calling” the outcome of the election early may inaccurately forecast a Trump win. The argument is based on sampling theory, and it goes something like this.
· Biden voters are more likely to use mail-in voting.
· Trump voters are more likely to vote in-person.
· In Wisconsin, in-person votes will be counted first. Why? The law says we cannot start counting mail-in ballots until election day, and it’s going to take a while to plow through the mail-in votes that have piled up over the past few weeks.
· Therefore, early forecasts could inaccurately predict a Trump win.
It all comes down to the idea that neither sample of voters, i.e. the mail-in sample nor the in-person sample, is representative of all the voters in Wisconsin. So, we wait…
More importantly, as it pertains to evidence-based practice, consider Wisconsin the next time you read a clinical research study. Was the sample representative of the entire population of interest? If the answer is yes, you’re golden! You can apply the results broadly. If the answer is no, then you must determine which subset of the population is best represented by the sample and limit your application of the data to those patients and clients.
Take care. And if you are reading from Kentucky, Illinois, or another state where the outcome is crystal clear, show your colleagues in Wisconsin a little grace today. We’re busy chewing our nails as we await the results of our study.
If you like what we’re doing, please consider supporting our work. Buy us a coffee - https://www.buymeacoffee.com/ptworkshop.
Copyright (2020) The Evidence Workshop, LLC. All rights reserved.