When you post, please see #11 in http://community.tableau.com/docs/DOC-5065 to make sure the post goes into the general forums where it will be visible.
I took a look at this and it's a head-scratcher. Here's what I found:
Create a new worksheet, SUM(Number of Records) on Columns, turn on Labels.
Put Sector of Institution on Filters Shelf.
Filter to Public only, see 362 records, Private only 618 (just like you did).
Create a new worksheet, SUM(Number of Records on Columns, turn on Labels, *add* Sector of Institute on Rows. See 1206 Private, 572 Public.
Put Sector of Institute on Filters Shelf.
Filters don't change results.
Then I started looking at the logs, pressed F5 a couple of times to refresh the view, and now the 1st view that was incorrect is returning correct results, and I can't duplicate the original error.
My first thought is to check whether you get the same problem when using a live connection vs. an extract, it might be something that only occurs for an extract. Either way, I'd suggest contacting Tableau tech support.
Thanks, Jonathan. The sound you hear is the relief that I'm not totally incompetent...
And thanks for the tip on posting. Never had the need to post here (at least that I recall) although I did check the box to show it to the whole community, I thought.
After I closed and reopened it, all the views disappeared. I went to the filters for SAT and ACT and saw they were red, rather than green. I pulled them off the pages, and everything works. I must have done something wrong with the configuration of the filters, but can't say what that might be right now. But at least I have a suspect.
The Jive software that powers the forums was upgraded, just ignore that "chose where to post" selection and you'll be all set in the future.
The more I think about this the more I think it's a problem with Tableau, because there shouldn't be any differences whether a the dimension for a single filter is in the view or not.
Please file a bug with Support. I can't find the workbook you originally linked to, so I don't have any insight into the problem.
Is it possible that you updated the extract (even just to optimize it) at some point but failed to save the workbook file? This can sometimes lead to changes that result in fields turning red, or fields with named like [Calculation_NNNN] suddenly appearing, typically due to extract optimization introducing materialized data for calculated fields.
Hi Robert. I think this is solved. I deleted the workbook and republished a new one.
Here is what I think happened: Some colleges in the set do not have ACT or SAT scores reported; some schools that have essentially open admissions are not required to. But I wanted people to be able to filter by mean ACT or Mean SAT CR+M scores. I calculated these based on the published 25th and 75th percentiles as reported. Everything was working fine. I pulled both to the filter shelf, then created quick filters. Everything still fine. Filtered on several other variables. Still good. But as soon as I activated the SAT or ACT score filter, things got funky.
I believe it had to do with a user mistake (mine, of course) in not understanding how null values figured into the filters. I noticed that when I re-set the filters, they would not go to zero (the original setting) but rather the minimum value in the data, which would exclude many of the institutions. I set them to include null values and all seems to be working fine now.
The new workbook is here. Thanks to all for the online and offline help. Workbook: IPEDS 2012 Dashboards