I’ve posted a corrected version* of our InfoVis paper from last year: An Empirical Model of Slope Ratio Comparisons. In preparing the published version of the paper, we made a change in the parameterization of our space of slope comparisons to simplify the explanation of what we did. In doing this, I made a simple math error that resulted in us using the wrong mid-angles in our analysis. To see the difference, compare Figure 2 in the original and in the updated versions. The impact of the error is minor and doesn’t change our arguments or conclusions, but it required regenerating our plots and it slightly changed our model parameter estimates.
I’ve also posted R code which will reproduce our (corrected) analysis and figures. Along with the stimuli we released earlier, this should allow anyone to reproduce our analysis.
(*The irony of having to correct our paper which itself attempts to correct Cleveland’s earlier paper was not lost on me.)