Have you tried downloading 7.0, upgrading to 8.0, publishing to 8.0.4, then extracting from 8.1, and creating a second level parmeter in the proposed 8.1.4? That may fix the problem.
I just tried your exact steps in both 8.0 and 8.1. In both cases the parameter does nothing when the filter is removed. If you add the Top N Set to the filters the parameter works as you described. Have I misunderstood the issue?
I've attached a sample workbook created in 8.0 that shows this behavior. The example worksheet was created as I described and shows the desired behavior as you change the parameter.
The reason why the filter must be removed is shown in the second worksheet, which is just a copy/paste of the first worksheet with a few tweaks:
1. Add column grand totals
2. Right click the "out of set" and hide it (as opposed to excluding it which would remove it from the grand total)
3. Right click the set pill and uncheck "show header"
The result of this is that the "in set" items are shown as dictated by the TopN parameter. The "out of set" items are not visible, but are included in the grand total. The effect of this is my best mechanism to force Tableau to provide a "Top N and Others" option. I don't want to exclude the others because the grand total must be intact . But for presentation purposes, I need to restrict the size of the display given that the data set almost always follows the Pareto principle. Ideally there'd be a subtotal for "others" in there, but since it's not currently possible in Tableau having the grand total and the power of math is good enough.
Our previously published reports exposed this parameter to the user so that they could pick their own Top N. Current behavior of 8.1 now precludes this.
What's very odd to me is that if I edit the set and manually change the Top number, the behavior is exactly as expected. Somehow it just doesn't seem to have a refresh/recalc triggered when the parameter is changed. I know this is not general parameter behavior because I leverage them heavily and they do trigger the calculation expected in all of my other use cases.
TopNTest.twbx 25.5 KB