Control chart builder: why is LCL =Avg = UCL when a time axis is added?

[#1]

When building a run chart in control chart builder and I drag the parameter onto the y-axis, I get correct spec limits and standard deviation. As soon as I drag the time column on the x-axis (subgroup), LCL equals Avg equals UCL. Why? Is there a way to have the time/date on the x-axis and still get a result?

I've attached a screen record to demonstrate the issue. Note, too, that the h:min:s turn to 00:00:00 on the x-axis

By the way, if I go through Legacy Control Charts -> IR, it works, LCL<Avg<UCL and correct h:min:s on the x-axis

Kofi

[#2]

There is something wrong with the the values in the Time data column. JMP detects multiple observations for the same value and so it switches from a XmR pair of charts to a XbarR pair of charts. Change the chart type back to Individual and Moving Range.

[#3]

I rather suspect there's something wrong with the Control Chart Builder. 

When I try to create a IR chart through the Control Chart Builder, JMP tells me that at least 1 sample has sample size > 1, hence can't create IR, has to make XBar (and consequently doesn't allow me to change back to individual and moving range)

When I try to create a XBar chart through the Legacy Control Chart menu, JMP tells me that no sample has sample size > 1, hence it can't create XBar, has to make IR.

What now? Does any of my sample has sample size >1 or not? Control Chart Builder platform says yes, Legacy Control chart platform says no. That's a bug!

Simple as that.

My guess: Control Chart Builder can't read it's own date-time format. Sets all h:min:s behind the date to 00:00:00 and then, yes, plenty of observations at the same time (=day/date). JMP needs to fix this

[#4 · Accepted solution]

I suggest that you examine the Time data column. There are several ways to determine if there are multiple rows with the same time.

• Select Tables > Summary and put Time in the Group role.
• Change the modeling type to Nominal and launch Distribution with Time in the Y role.
• Change the modeling type to Nominal and launch Distribution with Time in the Y role.
• Change the modeling type to Nominal and launch Tabulate with Time in a grouping role.

What did you find?

And what version of JMP are you using?

[#5]

Hi Mark,

that's a good one. Turned out the data table had 1 day entered twice = double entries.

Still remains the question why one platform picks up the duplicates and the other doesn't...?

I'm using v15.1.0

[#6]

The legacy control chart platform requires you to pick your control chart first. You ask for an IR chart, you will get an IR chart, even if it is incorrect for the data. The time variable on your IR chart would only be a label, so you would see the duplicated date twice on the control chart. It does NOT check to see if dates are repeated. The legacy platform requires that your data be in time order for the plot to be correct.

The control chart builder does not require your data to be in the correct time order. It DOES check to see if the data make sense for the plot. It has to do this so that it can create the correct chart. So the control chart builder is actually an improvement over the legacy platforms. Just think of your situation. You may not have found the inadvertently duplicated date without the control chart builder alerting you to its presence.

[#7]

This also answers my other question why h:m:s on the x-axis is 00:00:00 on the graph of the builder (and not on the legacy graph). Because its not a label, but a true point in time, in this case midnight of the respective days.
Thanks. I really appreciate the explanation.

[#8]

CCB adds a ton of control over control charts that the legacy charts were missing. Now in 15 there are even alarm scripts.  I was worried about the transition form legacy to CCB initially, but its become so seamless that I can't image going back.