Handling missing data when plotting variability charts with multiple Ys and BY groups

[#1]

m1	m2	m3	m4	m5	m6	m7	m8	Device	Family
1	2	-2	-1	-3	2	-1.5	0.5	D1	F1
2	-2	2	3	2	-1	1	2	D2	F1
1.5	2	2.5	2	0.5	1	-1	-1.5	D3	F1
-1	1	-1	-1	-0.5	1	0.5	-1	D4	F1
2	-3	1	-2	1	-2.5	0.5	-1	D5	F1
-3	-0.5	-1	1.5	.	.	-1	1	D6	F2
2.5	1	0.5	1.5	.	.	-0.5	2	D7	F2

I am going to use this table as an example. Assume that m1-m8 are some test parameters of some devices. In the actual table, the number of test parameters queried may be different. I want to plot variability charts using JSL for those parameters across Device for different Family.

In JSL, I did

myVC = Variability Chart(
			invisible,
			Y(Eval(parameterlist)),
			X(:Device),
			By(:Family)
		);

I have grouped the parameters m1-m8 as parameterlist. However due to the missing data, the code would result in error when it tries to plot the charts for F2.

Is there a way to handle the missing data? I want JMP to still plot the variability charts for F2 for the parameters that have data. (m1,m2,m3,m4,m7,m8 in this case). After that I want JMP to display the variability charts by family and save it as a html report.

F1 - m1

.

.

F1 - m8          (Order in which variability charts are shown in the report)

F2 - m1

. 

.

F2 - m8

Note that the number of test paramters, number of device and the number of different families may vary depending on the queries. 

[#2 · Accepted solution]

XiangLoh,

You might want to review an earlier post

https://community.jmp.com/t5/Discussions/JSL-Generate-x-y-Plot-that-loops-through-columns-with-diffe...

The recommendations, I made there, I would apply to your data. Here are the steps/ pseudo code:

• create parameterlist (already done)
• Run, summarize( fgrp=By(:Family) ); the fgrp is the list of family values
• nw = New Window("Tests",  vl = VListBox());

for( i=1, i<=nitems(fgrp), i++,
  fam = fgrp[i];
  for(j=1, j<=nitems(parameterlist), j++,
    param = parameterlist[j];
    idx = dt << get rows where(:Family == fam);
//    compute the number of non-missing values for column(param)[idx]
//    if that number is greater than 0 or some minimum you set then
      myVC = dt << Variability Chart( invisible, Y ( AsColumn(param) ) , X(:Device), By(:Family), where(:Family==fam) );
      get rid of the where where textbox and append to VL
); //end j
); //end i 

This method gives you a lot of flexibility.  You could save a different HTML for each family, or change the order and plot by variable then family.  Instead of the code for the 2 commented steps, you could use a Try()

Try(  myVC = dt << Variability Chart( invisible, Y ( AsColumn(param) ) , X(:Device), By(:Family), where(:Family==fam) );
//      get rid of the where where textbox and append to VL
,  //else do nothing 
   Show( "variable " || param || " for Family == " || fam || "has no valid data." )
 );   

It looks like a lot of work, but I have found that most scripters after they have done this once, and realize the amount of options this method offers, they use it often.

Good luck!

[#3]

Nice! Although I noticed that AsColumn in the variability chart doesnt seem to work for me. Changing it to Column (parameterlist[j]) works