It’s Not Statistically Valid

You Had Me At..

What does it mean to be statistically valid?

Let’s say you are in the first grade and your teacher wants to know what ice cream would be best to get you and your classmates.  Let’s say she asks a girl who says Ube  is what she should get for you and your classmates.  

Of course, most kids may not have heard of that flavor. Some may know of it by the nickname: the purple ice cream.  

The teacher now brings the class purple ice cream.  That girl would have reported back an awesome day at school.

If everyone let the teacher know what their favorite flavor was, what would the teacher have bought?

Who’s to say.

Anyway, let’s hit rewind.

If the teacher had everyone vote for their favorite, let’s say that 5 say vanilla, 2 cookies and cream, 7 chocolate,  6 for coffee, 5 strawberry, and one each for ube, Pistachio, and some blue ice cream.

If we were to put it on a graph…

On a bell curve, the chocolate and coffee would have the most on the chart.  Since the teacher only asked one student, her preference could have been where the majority would have been or, in this instance, not.

There is a population sample size where, if you get that number (minimum requirement), you can be confident that your feedback is, for example, 95% certain it represents the popular vote.  I won’t get into the specifics here on the math behind it, maybe another day.

Knowing this, makes me question every news voting results . Who is voting here, how large is their sample size, what’s their confidence interval, and why can’t I get a survey to take too?  I don’t think survey results should ever be published without any of that data shared.  ( It’s as credible as those Nielson ratings – who even has one of those boxes anyway? )

I once sat in a supplier review meeting, where they toted their high customer satisfaction ratings until I asked how many people they needed for their results to have a 95% confidence interval.   They had to take that answer offline.

Anyway, I think we need to go back to the popular vote and back to basic math and minimum requirements for voter response rate. Even if we don’t have a minimum response rate, it would be more of a diver for voters to vote.   (Once you start introducing a middle layer (delegates), how do we even validate that?!?)

If we had a popular vote, you’d have a say in the ice cream you get.  It’s the difference between coffee ice cream and purple ice cream.

Do it for the coffee.