Wednesday, July 15, 2020

Talk About A Small Sample Size!

"4 out of 5 dentists surveyed recommend sugarless gum for their patients who chew gum."

If you're "of a certain age", you recognize that line as being from a Trident gum commercial.  I'm sure more than a few of us wondered how many dentists they surveyed, if that number was only 5, and if so, why the 5th dentist didn't recommend sugarless gum.  Some of us were critical thinkers before that was even a thing.

Here's another small sample size, agree with the doctors or not:

In an endorsement of the Trump administration’s bid to reopen schools next month, 5 out of 5 pediatricians said they would return their own young children without hesitation.

“Yes. Period. Absolutely,” Dr. William Raszka of Vermont told NBC’s Dr. John Torres in a report late Sunday...

“The five doctors we spoke to agreed: The benefits of being in the classroom far outweigh the risk of disease. But the key is to reopen safely,” said Torres...

When Torres asked the doctors if they would let their kids return, here’s what they said:

  • Dr. Yvonne Mondonado, California: “I would let my kids go back to school.”
  • Dr. Shilpa Patel, New Jersey: “I will, my kids are looking forward to it.”
  • Dr. William Raszka, Vermont: “Yes. Period. Absolutely.”
  • Dr. Jennifer Lighter, New York: “Absolutely, as much as I can. Without hesitation, yes.”
  • Dr. Buddy Creech, Tennessee: "I have no concerns about sending my child to school in the fall.”
Let's talk about confidence intervals.  When we use sample data to estimate a proportion (e.g., what is the true proportion of all pediatricians would would send their kids back to school?), we generate what's called a confidence interval.  When we generate what's called a "95% confidence interval", that means that our sampling is such that if we were to randomly sample our population (in this case, pediatricians) 100 times, 95 of the confidence intervals created would contain the true proportion that we're studying (in this case, the percentage of pediatricians who would send their kids back to school).  Using the data above, though, a 95% confidence interval would not be an interval at all--it would be the number 100%.  Thus, based on this small sample with a sample proportion of 100% (5/5), our confidence interval would be [100%, 100%].  This means that the true proportion of pediatricians who would send their kids back to school is between 100% and 100%--again, not much of an interval.  It's certainly not believable and is not an example of a good use of statistics.

I suggest a larger sample size!

1 comment:

Ellen K said...

On our local Fox affiliate a very earnest reporter, socially distancing by reporting from her Plano home, asked the Chief Medical Officer of Parkland Hospital (one of the largest ER hospitals in the nation) if he would let his own children attend school in person. His response was "My children are teenagers and their age group learns better with peers. I feel it is necessary for their benefit for my girls to go to school in person."

The reporter, who is also the wife of a pediatrician, was visibly unhappy with his response.