In the attached illustration, I have plotted the Average predicted value of my statistical model against 10 buckets of one particular independent variable.
I then tried having each bar discretely shaded in by the percentage of actual predictions vs. non-predictions 0 vs. 1. One specific color for 0s and another for 1s. I played around with several options for a while but the tool kept creating separate average predicted value plots for 0 and 1.
I tried to re-route my strategy and create a calculated variable, specifically what percent of each bucket was wins. The only effective thing that could be done with this is a dual axis or color intensity for each bar for higher percentages.
What I am trying to actually achieve is to split the bar for each bucket by the numbers of 0s vs 1s. Displaying that on an axis is irrelevant. I want to be able to visually see it. For example for the first bucket $(559) - $49.99 , if the number of 1s is 75% of the total i want 75% of the bar shaded one color and the other 25% another color without any changes to the y axis (Avg. p).
Avg. P is the average probability of being a 1 as determined by the statistical model. So my goal here is to show, of that total bar size, what portion of it (graphically) was an actual 1.
Thanks.
I've spent hours trying to figure this out. Unfortunately the Tableau tool does not seem to have any quick drag/drop or click through tool to execute this process.
I worked around this by creating a calculated variable: (sum([WIN01])/COUNT([WIN01]))*(AVG([p]))
Then placing it into columns with dual axis on. This will shade in the percentage of the actual bar that are wins/no wins.