Process Capability Physical Boundaries

Goblin_King

How can one set physical boundaries in process capability analysis.  For example, when I evaluate capability to measure within specification of a trace element in a material, most data will be near zero, but cannot be less than zero.  This results in a highly skewed distribution with the bulk of the data piling up around zero and then the distribution has only a one-sided tail.  But the capability analysis has no way I can find to set zero as a physical boundary, so the Ppk value is always artificially low because the statistical fitted distribution assumes tail less than zero.  Is there some other way I can compensate for the lack of this feature.  I would never want in any way to praise Minitab over JMP, but Minitab does have the option to set the physical boundary and I cannot find for the life of me such an option in JMP.

tonya_mauldin

You should be using one of the nonnormal distributions if your data does not follow the normal distribution.  It's hard to say the appropriate distribution without seeing your data.  It sounds like perhaps the Lognormal distribution may work.  You can find an example of this in the Object Scripting Index.

dt = Open( "$SAMPLE_DATA/Process Measurements.jmp" );
obj =
Process Capability(
	Process Variables(
		:Process 1 & Dist( Lognormal )
	),
	Individual Detail Reports( 1 )
);
obj << {(:Process 1 & Dist( Lognormal )) <<
Process Capability Analysis(
	Compare Distributions( 1, <<Fit Lognormal )
)};