add test statistics ref line to visualize()?

[tidymodels]

add test statistics ref line to visualize()?

[andrewpbray]

Our most recent thought was to calculate the observed statistic in a separate pipeline and then include it as an argument to visualize(), I think.

[beanumber]

A separate pipeline? That seems way to clunky. Once the hypothesis is specified (wink), shouldn't we have everything we need to know? I'd favor having hypothesize() add a test_stat attribute before it craps out.

[ismayc]

I think the inclination was to have students/users use the strategies that they have developed with dplyr to calculate the observed test statistic in a different pipeline since this isn't too many links in a chain.

[beanumber]

Can you provide a full example of how you think this would work?

[ismayc]

Here are some thoughts and code in a gist. My interpretation of previous discussions. Completely open to other ideas here.

[ismayc]

Until we have both computational and approximation methods implemented, I think it best that students write their own ggplot2 code instead of using visualize to produce the histogram/bar graph and vertical line. It also gets them in the habit of saving the chain from hypothesize to calculate and assigning a name like null_distn. They can then use the null_distn and one line of summarize code to get the p-value, depending on what H_A is:

null_distn %>% summarize(p_value = mean(stat <= obs_stat))

Thoughts?

[mine-cetinkaya-rundel]

I like this approach. I thought I added a comment here earlier, clearly not. I had meant to say that it doesn't seem worthwhile to wrap a ggplot call, or a ggplot + geom_vline code in another function. If we want to do shading, on the other hand, that might be code we don't want to confuse students with.

[andrewpbray]

The main reason for the wrap would be to accommodate approximation methods. My sense is that if they were to recreate the same ggplot on their own, the syntax would have to look pretty different and some of the similarities between the methods would be lost.

[ismayc]

@andrewpbray Can this be closed?

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.