Bias

 

That Song About the Midway

So lately you've been hiding, it was somewhere in the news
And I'm still at these races with my ticket stubs and my blues
And a voice calls out the numbers, and it sometimes mentions mine
And I feel like I've been working overtime, overtime

Do the numbers ever lie?

I have had to conduct, or use, numerous surveys in my engineering career.  Surveys are done because it is NOT cost effective to observe the behavior of an entire universe. However you can obtain statistically accurate results if the sample being surveyed looks like the universe.

If the survey sample looks like the universe, then the sample is said to be unbiased.  If the sample does NOT look like the universe, then the sample is said to be biased.  ( yes I know that this is an emotionally loaded term, but that IS the statistical term). That does not mean that the sample is invalid, just that the portion that is underrepresented has to be oversampled in order to eliminate the bias.

For example, if according to the 2020 Census, the US population is 13.4% black and 50.8% female then in a nine person sample, say SCOTUS, there should be 4.5 women and 1.2 blacks.  If there are 8 justices on the Supreme Court and only 3 of them are women and only 1 justice is black, then the next justice on SCOTUS should be a black woman.

It does not matter if the sample is screws, household travel, corporations, or Supreme Court Justices.  If the sample is biased then a method to correct that bias is to oversample.  If you oppose that oversampling, then by definition you are biased. Figures don’t lie, but liars do figure.