So one way we can do this is to use Table Calculations to work out the Whiskers of the Box Plots, and then we can test if a customer has 1 (or more)...
First I had to find out the exact rule Tableau uses for it's Box and Whiskers plot...luckily it's all detailed here https://www.theinformationlab.co.uk/2017/05/03/box-plot-calculations-tableau/
So I created these formulas and set them up as per the below (please note the sort)...they all have the same Table Calculation set up
Once we have this I can test each mark against the Top and Bottom Whiskers
[Customer Outlier 1/0]
IF SUM([Sales]) > [IQR: Upper Whisker]
OR SUM([Sales]) < [IQR: Lower Whisker] THEN 1 ELSE 0 END
and then finally I can use a WINDOW_SUM to find which customers have 1 or more
[Customer Has Outlier T/F]
WINDOW_SUM([Customer Outlier 1/0]) >0
This one is just set up like (same as rest, but no need to have the sort...it wouldn't make any difference, but why work harder!)
We can now use this as a filter.
Hope that makes sense, and helps (attached in 10.4)
IQR Box Plots and Outlier Filters.twbx 1,012.5 KB
Thanks for the help. However, I noticed that it does not work fully.
For example looking at Sarah Middleton, it gave a different outlier result according to number of customer selected. From your attached file, without changing any item, Sarah Middleton's hinge & whisker calculation does not match the Tableau box plot value. Even though there's no outlier, the outlier condition still return "true". However, when I "Keep Only" Sarah Middleton, the hinge & whisker matches Tableau box plot and it returned "false" for the outlier condition.
Do you have any idea why this is happening?
Apologies...I should have done a bit more checking!
I've had a look and all the calculations need to be set up with this arrangement (you can drag Customer Name to change the order with Month, so you can select it in the Restarting every)
Hopefully that does the trick.
IQR Box Plots and Outlier Filters.twbx 1,022.1 KB
Thanks for the trick! I'm able to get more accurate results. However, I do noticed that the Upper Whisker is still incorrect for some.
Example for Joan Oakley Schaefer, her Upper Whisker calculation does not meet Tableau box plot value. However, all her other calculation matches. Similarly, when I "Keep Only" Joan, her Upper Whisker calculation will match.
Any idea why this is happening?
Apologies for not getting back sooner
...looks like we're getting closer!
So I have slightly amended the Table Calculation set up for the Whiskers, so the Upper and Lower Whiskers are now like this
and the Hinges are as before
This has been and interesting problem, and I've learned a lot here...which is what the community is all about!
If you spot any other anomalies with Tableau vs my Calculation let me know
Sorry for the delay... I was having some issue correcting my own dataset.
I don't see any more issue in your attached workbook. Thanks for all the help you rendered. It really helps me a lot!