Iteratively Create Columns Containing All Permutations of Subtraction from an Initial List of Columns

[#1]

I have a list of columns and I need to create a new series of columns with every permutation of the difference between the listed columns.  The number of the columns in the list can vary and the non-iterative example below covers 3 columns (12 to 14).

Current  Data Table() << New Column(Head Name( As Namespace( Column( 12 ) ) )||" - "||Head Name( As Namespace( Column( 13 ) ) )||" HSD Threshold Matrix", Numeric, "Continuous", Format("Best", 10), Formula(Col Maximum(If(Contains(:Level, Head Name(As Namespace(Column(13))), 1), As Column(12)), :Group, :Size)));

Current  Data Table() << New Column(Head Name( As Namespace( Column( 12 ) ) )||" - "||Head Name( As Namespace( Column( 14 ) ) )||" HSD Threshold Matrix", Numeric, "Continuous", Format("Best", 10), Formula(Col Maximum(If(Contains(:Level, Head Name(As Namespace(Column(14))), 1), As Column(12)), :Group, :Size)));

Current  Data Table() << New Column(Head Name( As Namespace( Column( 13 ) ) )||" - "||Head Name( As Namespace( Column( 14 ) ) )||" HSD Threshold Matrix", Numeric, "Continuous", Format("Best", 10), Formula(Col Maximum(If(Contains(:Level, Head Name(As Namespace(Column(14))), 1), As Column(13)), :Group, :Size)));

[#2]

You can for example use NChooseK Matrix() to create the combinations (or DoE)

Names Default To Here(1);

mylist = {12, 13, 14};
possible_group_idx = NChooseK Matrix(3,2);

my_groups = Transform Each({idx}, possible_group_idx,
	mylist[idx]
);

[12 13, 12 14, 13 14]

• How to randomly assign one value to three unique IDs, while having each ID assigned 6 unique values?
•  Combinations of a list put into columns 
• JMP can generate all possible combinations of 40 numbers in groups of 6 numbers without repetition? 

[#3]

Jarmo

Thanks, looks like it will work.  Haven't had time to study the details but wondering if there's an example of how it can be employed?

[#4]

Loop over the matrix (or turn it into list of lists) and create the columns as you do that.

[#5 · Accepted solution]

Not sure if this is what you meant but I found something here that works.  What's even better about this is that it doesn't rely on column numbers.

Names Default to Here(1);
dt = Current Data Table();

//Select X Variable ____________________________________________________________________________________________________________________________________________

cdb = Column Dialog("Add continuous variables you want to plot as X" ,
	colxx = ColList( "X", Min Col( 1 ) ),
);
show(cdb);
Eval List(cdb[1::N Items(cdb)-1]);

Summarize(dt, uniqNames = by(colxx[1]));
Show(uniqNames);
Num = NItems(uniqNames);



For( k = 1, k <= N Items( uniqNames ), k++,
	// Loop through all columns finding non NA values
	For( i = 2, i <= N Items( uniqNames ), i++,
		if (i > k, Current  Data Table() << New Column(Char(uniqNames[k])||" - "||Char(uniqNames[i])||" Matrix", Numeric, "Continuous", Format("Best", 10), Formula(Col Maximum(If(Contains(:Level, Char(uniqNames[i]), 1), As Column(uniqNames[k])), :Group, :Size)));
			)
		)
	)
;

Solved: Get column values using Loop - JMP User Community

[#6]

Names Default To Here(1);
dt = Open("$SAMPLE_DATA/Cities.jmp");

mycols = {"OZONE", "CO", "SO2", "NO", "PM10"};

possible_group_idx = NChooseK Matrix(N Items(mycols), 2);
my_groups = Transform Each({idx}, As List(possible_group_idx), Output("List"),
	mycols[idx]
);

For Each({cols}, my_groups,
	Eval(Substitute(
		Expr(dt << New Column(cols[1] ||"_"||cols[2], Numeric, Continuous,
			Formula(_col1_ + _col2_)
		)),
		Expr(_col1_), Name Expr(AsColumn(dt, cols[1])),
		Expr(_col2_), Name Expr(AsColumn(dt, cols[2]))
	));
);

Write();