Okay so I did a simulation to show what happens. I simulated a perfectly calibrated data set and then thinned the 1 class to be 9.5% of the entire dataset. This is the calibration plot.
This is a better version IMO but still I want to consider something besides a heatmap and of course the colors on this aren't the best.
trying a heatmap for visualizing this clustering of a similarity matrix but there's got to be a better way, because in reality i have more than there 14 groups of 7, I have like over 100 groups of 7. The labels on this example aren't great, but the graph gives you an idea of what you're looking at. #rstats #dataviz
Data scientist at The Router Company. PhD in statistics at iastate, ex Indiana U visiting Asst. Prof, focusing on statistical computing, specifically clustering and other unsupervised or semi-supervised problems.