Question 1

For each of the studies below, explain the sort of bias or biases the study or information could be prone to.

1. You want to know how common mental illness is among the homeless in your county. Over the course of a week, you conduct mental illness tests among samples of those at all the homeless shelters around the county.
2. You want to know how common it was for felons to have been abused as children. You go to a prison, announce you are conducting a study on the childhood experiences of prisoners, and then conduct interviews with those willing to participate in the study.
3. The PTA organizes after-school tutoring to help kids with their homework. After six months they find that students who took advantage of this help had 25 percent higher scores than did those who did not take advantage of the help.
4. You read an ad: "Four out of five dentists recommend Bright toothpaste over the nearest competitor."
5. Random telephone calls were made to find out if citizens would vote for President Trump for President if the election were held today.
6. To understand the effect of educational attainment on crime, you gather information on the educational attainment of those convicted of crimes, and compare this with those who have not been convicted of crime. The results are striking, and you use them to advance the case for more spending on education.

Question 2

Make up your own examples of how data are generated (like those above) that you think might be prone to each of the biases discussed in class (provide one example for each): a) selection bias, b) publication bias, c) recall bias, d) survivorship bias, and e) healthy user bias (also called "omitted variable bias" or spurious correlation). Label each example a) through e), and be sure to explain why the example shows the given bias.