I had a hard time coming up with a candy activity for bivariate data that was really good. I ended up kind of cheating, and wrote an activity that has students gather lots of data points, one of which is “how many jelly beans can you pick up in one no-palm pinch?” and then letting them check the correlation of that with any of a wide variety of data points (height, hair length, finger length, age in days, etc.).
I don’t think it is as good, or will allow for quite as deep a discussion, as the earlier activities, but it is still decent I think.
I still haven’t actually taught regression inference, so I’m using this activity right after a brief mini-lesson on interpreting and using computer output for regression inference. I definitely will not be tackling this topic at the depth our book does; will focus instead on practicality and the basic idea of extending our knowledge of T-tests and confidence intervals to a more uncharted area. Some of the data sets they decide to plot will probably NOT satisfy all of the conditions for regression inference, though, so that could lead to a good discussion.
You can find all of my AP Stat candy activities, including this one (forgive any typos, it was made very late last night) here: https://drive.google.com/open?id=0B-C-lUvv4rQ4ZG9hcnh1azdHNjQ&authuser=0