Let Process Screening groupings include missing or blank values, maybe optionally

[hardner]

Say you do the following, slightly modified from the scripting index:

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Semiconductor Capability.jmp" );
obj = dt << Process Screening( Grouping( :Site, :lot_id ), Process Variables( Eval( 5 :: 132 ) ), automaticrecalc(1) );

You'll get a report like this with rows for the different combinations of values from the Grouping() columns:

And those rows act as a gateway to various plots.  For example,  you can get Process Capability or Control Charts that are pretty much as if you had used these grouping() columns as a By() on those platforms.  But you won't get a row for a grouping where any of the Grouping column values were missing or, if character, whitespace.  So if you replace a lot_id value with blank or a SITE with missing you don't see that appear on the report so you don't know there was some data that was like that, maybe needing attention.

When I discussed this with JMP support I was told that this is consistent with other JMP platforms, for example how X values that are blank don't show up on the X axis of a Oneway.  But my perspective is that Grouping() is a lot more like a By() in this case (which does make plots for the by groups with values of missing or blank).  The situation with Grouping() is also similar to summary/summarize which would recognize those groups and produce a row.

I appreciate that Process Screening has a By() and Grouping() is not it, but I think there would be value in showing up these groups and providing access to the same plots you'd see from a by()  (ie if you ran Control Chart() with a by() on the same columns you would see a report for a missing SITE):

The upshot is that I was encouraged to after all make a Wish List item on this so this perspective could be considered. 

I think it would at least be worthwhile as an option.

Why is this idea important? 

The main issue/value is how starting from Process Screening as your gateway to the various other reports lets you miss some of your data or problems in your data.  This came to my attention while working with a customer with a large amount of data in many groupings with many levels per grouping and with data that comes from machine sources.  To some degree the Process Screening is advertised at letting you sift through lots and lots of processes and so it's a situation where the possibility that some data is defective - and that that could go unnoticed - is heightened.  This platform could even go so far as to highlight such groupings, let alone include them.

[jthi]

I think most of the platforms in JMP should support this in some way (not just Process Screening). For example Graph Builder already does and so does Neural Network when using JMP Pro.

Graph Builder has option to treat missing as separate category

And Neural Network can do same (with JMP Pro)

If this setting by default is too confusing, it maybe be hidden behind the shift+click on red triangle, so no one will ever know about it (like showing excluded rows in graph builder).

[hogi]

@hardner , @jthi, 

[karenb]

I have the same concern with the Process Screening platform. I thought that all of the data was summarized in our complex set of BY variables and only later realized that we had not reviewed and approved all of the control charts.  I hope that this can be implemented.  We look at it as a gap in the BY-group processing that is hard to remember and hard to caution people to avoid.

Thanks