Exercise

# When the null is true

In this exercise, you will run an experiment: what happens when you conduct a hypothesis test when you *know* that the null hypothesis is true? You hope that you will retain the null hypothesis, but there's always a chance that you will make a statistical error.

To begin the experiment, we have created a new explanatory variable called `coinflip`

that captures the result of a fair coin toss for every subject. With that variable in hand you can pose the following null hypothesis:

$$ H_{0}: p_{heads} - p_{tails} = 0 $$

This claims that there is no difference in the proportions that favor the death penalty between the people that flipped `"heads"`

and those that flipped `"tails"`

. Since `coinflip`

was formed independently of `cappun`

, we *know* that this null hypothesis is true. The question is: will your test reject or retain this null hypothesis?

Instructions 1/4

**undefined XP**

*Inspect the new*`coinflip`

variable.- Compute the proportions that
`FAVOR`

among both`heads`

and`tails`

values of`coinflip`

using the dataset`gssmod`

. Save this statistic to`p_hats`

.