MantisBT - ParaView
View Issue Details
IDProjectCategoryView StatusDate SubmittedLast Update
0014886ParaView(No Category)public2014-07-24 19:512016-08-12 09:59
ReporterKen Moreland 
Assigned ToKitware Robot 
PriorityhighSeverityminorReproducibilityhave not tried
StatusclosedResolutionmoved 
PlatformOSOS Version
Product Versiongit-master 
Target Version5.1Fixed in Version 
ProjectSandia
Topic Name
Typeincorrect functionality
Summary0014886: Descriptive statistics max appears wrong
DescriptionThe descriptive statistics filter seems to be providing the wrong value for maximum, at least when it is used in conjunction with plot selection over time. To replicate, do the following:

1. Load can.ex2. Turn on EQPS. Apply.
2. Merge blocks. Apply.
3. Add Descriptive Statistics filter. Change attribute mode to Cell Data. Select EQPS variable (only). Apply.
4. In the spreadsheet of the statistical model, select the row associated with EQPS. (If you only selected the EQPS previously, it should be the only row.)
5. Add the Plot Selection Over Time Filter. Apply.
6. Delete the spreadsheet views (to see the plot more easily). Select only the Maximum series. Notice the weird sawtooth nature of the plot. That is not right. It should be mostly (if not totally) monotonically increasing.
7. Select the Merge Blocks filter in the pipeline browser.
8. Open Find Data. Query for the max EQPS. Plot selection over time. Apply.
9. In this plot, select only the maximum EQPS value. This curve shape is expected.

I've attached an image of the results.

The two versions of maximum should give the same value. Considering how weird the curve from descriptive statistics looks, I believe that is the one that is incorrect.
Steps To Reproduce
Additional Information
TagsNo tags attached.
Relationships
Attached Filespng ParaView002.png (257,564) 2014-07-24 19:51
https://www.vtk.org/Bug/file/9683/ParaView002.png
png
Issue History
Date ModifiedUsernameFieldChange
2014-07-24 19:51Ken MorelandNew Issue
2014-07-24 19:51Ken MorelandFile Added: ParaView002.png
2014-07-24 20:37Alan ScottProjectTBD => Sandia
2014-07-24 20:37Alan ScottNote Added: 0033081
2014-07-24 20:37Alan ScottPrioritynormal => high
2014-07-24 20:37Alan ScottTarget Version => 4.3
2014-07-25 19:51David ThompsonNote Added: 0033096
2014-07-27 17:47Ken MorelandNote Added: 0033106
2014-11-14 22:52Utkarsh AyachitTarget Version4.3 => 4.4
2015-09-11 16:43Utkarsh AyachitTarget Version4.4 => 5.1
2016-06-10 10:55Cory QuammenNote Added: 0036057
2016-08-12 09:59Kitware RobotNote Added: 0038607
2016-08-12 09:59Kitware RobotStatusbacklog => closed
2016-08-12 09:59Kitware RobotResolutionopen => moved
2016-08-12 09:59Kitware RobotAssigned To => Kitware Robot

Notes
(0033081)
Alan Scott   
2014-07-24 20:37   
This is misrepresenting data, thus I am targeting it for 4.3.
(0033096)
David Thompson   
2014-07-25 19:51   
Since no one mentions adjusting the default parameters of the Descriptive Statistics filter, I assume this is caused by training the model using only half of the dataset (the default). This was chosen so that people would evaluate statistical models using the remainder of the dataset rather than blindly supposing the data fit the assumptions of the model.

If people are using the filter differently, then either
1. the filter should be modified so that a minimum sample size is enforced (at least 100,000 samples) or
2. the default mode of operation should be changed.
Please do not just change the default training set fraction to be 1.0.
(0033106)
Ken Moreland   
2014-07-27 17:47   
That's a good assumption. I suspect most people are using the descriptive statics filter simply to get the model information (average, min, max, stddev). In fact, I would venture to say that most users don't even know how to use the model to evaluate (myself included).

I would suggest changing the default behavior to base the model on all values. Although sampling should give you a reasonable value for average and standard deviation, it's highly likely to fail to get reasonable values for min and max. Min and max are just wrong (statistically or otherwise).
(0036057)
Cory Quammen   
2016-06-10 10:55   
If you want the min and max for the entire data set, you can change the "Task" option to "Detailed model of input data". The documentation for this in the filter is fairly clear.

We could change the default Task type to this if you would like.
(0038607)
Kitware Robot   
2016-08-12 09:59   
Resolving issue as `moved`.

This issue tracker is no longer used. Further discussion of this issue may take place in the current ParaView Issues page linked in the banner at the top of this page.