Control chart, alarm and missing values

Question

Hi,

Following my previous post Control chart, alarm and filter I have now my data who are splited.

But, it generates me some missing values (and sometimes, in your data, you have missing values, without my data manipulation)

I create a control chart with warnings.

In this case, I activated ALL the warnings. But we can see that the test 2 doesn't apply (due to the fact it considers that data are not consecutive. But for this variable, it is. I just have missing values due to sensor frequency or whatever!!!)

is there a way to avoid that?
As you can see I need a column switcher so removing or sorting the data is not a solution!

Thank a lot for all your suggestions!!

Florence

AnnieDudley · Answer

Also see Tonya's description of how Show Excluded Region() offers a workable solution.

Control chart, alarm and filter

AnnieDudley · Answer

Hello Martin and Florence,

Thanks for reaching out for additional help. As we discussed on Thursday on our phone call, I believe the only approach that will work in this case is using a "By"-variable. I understand it's not ideal, as each By-group becomes a separate data table/output and needs to be manipulated separately, however that's the best approach I know currently. Here's an example script:

dt = Open( "$SAMPLE_DATA/Semiconductor Capability.jmp" );

Control Chart Builder(
    Variables( Subgroup( :Wafer ID in lot ID ), Y( :NPN1 ) ),
    Chart( Position( 1 ), Warnings( Test 1( 1 ), Test 2( 1 ) ) ),
    Column Switcher( :NPN1, {:NPN1, :PNP1, :PNP2, :NPN2, :PNP3, :IVP1, :PNP4, :NPN3} ),
    By( :SITE )
);

Some background on why the tests are working the way they are: Around 20 years ago, I met with users, Testing, and Tech Support to determine how traditional control chart users need the tests to work. The users suggested the current rules, which are documented in the "Excluded and Hidden Samples" section of the Control Chart Builder chapter, see below. In the section, you'll see twice mentioned that 'Tests apply to all excluded subgroups'.

https://www.jmp.com/content/dam/jmp/documents/en/support/jmp151/quality-and-process-methods.pdf

martindemel · Answer

Florence,

if you use the data filter e.g. for one of your columns you use the column switcher for and is active. Select in the filter hotspot for this variable "Select Missing", then check "Inverse", the points will be connected and the gaps will be removed.

However, as in your previous thread, it seems the tests will be calculated on th ewhole column no matter if you filter or not. Which means that missing values or values from a different group level will still be counted (the latter probably as missing). And treated as all but not on the same side (for test 2 e.g.).

I believe the test filtering is wrong but there might be a reason I just don't know of, why you still want to use the whole data for calculating the tests even if you have excluded the data. @AnnieDudley may know more?