In#361 and#358, as part of our discussion related to enhancing error/residual plots in general, we talked about implementing a new function namedppc_residual_scatter. The idea of such a function originated in#349 and#350 due to a demand for a function that creates plots ofy - stat(yrep) instead of what existing error functions plot,stat(y - yrep). So, the logic ofppc_residual_scatter is what@jgabry said in his commenthere:

2. Keepppc_error_scatter_avg as you have it and add a new one that doesy - stat(y_rep) on the y-axis andstat(y_rep) on the x-axis. We could call itppc_residual() or something like that. I think with the right documentation having both would probably be ok.

This PR is about my work on creating theppc_residual_scatter.

• implementppc_residual_scatter
• write documentation
• add tests

The problem I have is that I am a little confused about my implementation. When I look at it, it seems logical and in some cases, I am fine with the results. However, there are some cases which confused me, such asppc_residual_scatter(y, yrep, x) andppc_error_scatter_avg(y, yrep, x) having the same plot. I think these plots shouldn't look identical, but somehow, to me, they look like they are the same plot.

Examples

y<- example_y_data()yrep<- example_yrep_draws()x<- example_x_data()ppc_residual_scatter(y,yrep)

ppc_error_scatter_avg(y,yrep)

ppc_residual_scatter(y,yrep,x)

ppc_error_scatter_avg(y,yrep,x)