Q (Gelman):

I don't understand what's going on. What's the x and y-axis?

A:

Once again, the x-axis is just the score of the demographic space, so that's my predicted score.

Q (Gelman):

The x-axis is labeled Bayes prediction, not expected value of the data.

A:

Remember what the model is, it's is equal to plus gamma demographics plus random error, so what I want to do -- I've got twelve demographics, or eleven demographics here, so I could show you eleven plots for eleven demographics variables, and instead what I'm doing is I've just given you a plot of against this gamma demographics. The point is I'm just trying to regress it, my y against my . But the least square-estimates don't necessarily have to have any relationship, so my least-square estimates are not , it's my Bayes estimates that are . Obviously you're going to see this relationship between the Bayesian demographics but it's not clear that your're going to see this with the least-square estimates in the demographics.

Q:

So the reason why the black diamonds fall near the 45-degree line is these store effects which are on the y-axis?

A:

It's these random effects that are going on. It's just a way of trying to illustrate what this stuff looks like.

Back