How to show failure rate in summary statistics (e.g. percentage of cells in a column with value larger than a given number)?

[#1]

For example, snapshot below is the statistics of a column of values. There are mean/Std Dev etc in the Summery Statistics. I want to add another variable called failure rate, which is defined as the number of cells out of the total N (=1079) which has value larger than a give number, e.g. 99.5. How can I do that? Thanks.

[#2]

I'm not sure if you can directly modify Summary Statistics with custom data, but you could create your own Summary Statistics with something like this:

Names Default To Here(1);
dt = Open("$SAMPLE_DATA/Big Class.jmp");
obj = dt << Distribution(Column(:Height));
obj << Summary Statistics(1);
a = report(obj);
b = a["Summary Statistics"][TableBox(1)];

//values from current summary statistics
vals = b << get; 
//add new text value
Insert Into(vals[1], "Failure Rate");
//get new num value
valueOver = N Items(dt << Get Rows Where(:weight >= 100));
//add new number value based on N index
Insert Into(vals[2], valueOver);
//create custom summary statistics
a << Append(
	Outlinebox("Summary Statistics Custom",
		Table Box(
			String Col Box("",vals[1], << Background color(80)),
			Number Col Box("",vals[2])
		)
	)
);
//hide original summary statistics
//obj << Summary Statistics(0);

[#3]

Use the distribution platform and turn on the summary statistics, fit a model to your data (normal for example) and then in the normal hotspot turn on capability, fill int he spec limits and you will get failure rate

[#4]

In addition to both other replies, I think this is best covered by the process capability platform already. So if you add a spec-limit to that column (i.e. USL=100 for the column weight in Big Class.jmp like @jthi did in his script), you will get in the distribution platform also a process capability, which shows you the Nonconformance shares (55% = 22 members of Big Class.jmp).

There you can also see the difference of real nonconformance to the modeled nonconformance share. The modeled one is only ok, when the model fits the distribution well.