danryan.us

Q103. Suppose we are running a program to which we want to accept only individuals in the top 25% of the population (on some measurable trait). Unfortunately, our test for measuring the trait is only 80% accurate. Draw event tree and flip to show what kind of faith we can have in the test results. Which test result appears more worthy of taking at face value? Which group would you be inclined to develop a second test for?

+ Show solution

- Hide solution

Again, we use our tree flipping tool. Imagine this is the Mensa test. What does it make you think about Mensa now?

Apparently the test does a good job at detecting the average dolt. On the other hand, it seems to provide a lot of folks with an overinflated view of themselves. What are the policy implications if this sort of thing were true of credentialing systems in general? How likely do you think it is?

File name	File type	Size
DT-top-quartile-test.xlsx	No description	16.45 kB	Info

Manage attachments