add correlation coefficient to the correlation plot (Trac #437)

[pkienzle]

For each plot showing a strong correlation (e.g., >0.2) between a pair of parameters, plot the correlation value on the plot. The documentation will need to include information about how to report errors when there are strong correlations. The correlations should also be included in the parameter output table, along with the overall correlation coefficient and other statistics.

Migrated from http://trac.sasview.org/ticket/437

{
    "status": "new",
    "changetime": "2017-10-23T15:42:18",
    "_ts": "2017-10-23 15:42:18.200371+00:00",
    "description": "For each plot showing a strong correlation (e.g., >0.2) between a pair of parameters, plot the correlation value on the plot.  The documentation will need to include information about how to report errors when there are strong correlations.  The correlations should also be included in the parameter output table, along with the overall correlation coefficient and other statistics.",
    "reporter": "pkienzle",
    "cc": "",
    "resolution": "",
    "workpackage": "SasView GUI Enhancements",
    "time": "2015-06-29T15:22:04",
    "component": "SasView",
    "summary": "add correlation coefficient to the correlation plot",
    "priority": "minor",
    "keywords": "",
    "milestone": "SasView 5.1.0",
    "owner": "",
    "type": "enhancement"
}

[pkienzle]

Trac update at 2015/06/29 15:23:36: pkienzle commented:

This ticket is in response to the following comment on the 3.1.0 release:

12. In the DREAM parameter 2d correlation plots are very good but don’t actually quantify how large the correlation is. It might be useful to have some scale markers on the axes. The 2d plots are of course very small, so normal axis labels would likely look a mess, perhaps try sticks on the sides at 0 and say +-3sigma. (Normal correlation coefficients as I recall say how many standard deviations parameter A moves if parameter B moves one standard deviation.)

PAK: yes, I left off the labels because the plots would be too busy. Since the points used on the correlations are the same 95% interval as used on the 1-D variable plots, you can use those plots to read off the actual values, or look on the console log to get a parameter table. Printing the correlation coefficient on the plots where there is significant correlation

[butlerpd]

Trac update at 2015/12/19 18:58:18: butler changed workpackage from "SasView Bug Fixing" to "SasView GUI Enhancements"

[butlerpd]

Trac update at 2017/10/23 15:42:18: butler changed milestone from "SasView Next Release +1" to "SasView 5.1.0"