<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />Hi Folks,
Im just wondering if there is anyway to write a script that pulls out the split candidates from a partition analysis in a data table, ordered by logworth. This can be done manually when you fit a partition model and right click on the candidates section.
Perfect David, just what I was looking for
Scott
I do this quite a lot. Here is anexample of the code that I use
( the answer to your question are the 4 lines in the section "extract the candidates table" ! )
// sample data
dt = Open("$SAMPLE_DATA/Fitness.jmp");
dtName = dt << Get Name;
// select response variable
colNames = dt << Get Column Names(string);
nw = New Window("Select Y", <
lb = List Box(colNames, MaxSelected(1), <(5))
);
If (nw["Button"]==-1, Throw());
sel = lb << Get Selected;
If (Nitems(sel)==0,
Throw()
,
yCol = sel[1]
);
// use all other columns as x variables
pos = Loc(colNames,yCol);
xCols = Remove(colNames,pos[1]);
// drop name column
RemoveFrom(xCols,1);
// launch the partition platform
pat = Partition(
Y( Eval(yCol) ),
X( Eval(xCols) ),
);
// extract the candidates table
rep = pat << Report;
candTab = rep["Candidates"][TableBox(1)];
candDt = candTab << Make Into Data Table;
rep << Close Window;
// sort the candidates table
candDt << Sort( By( :LogWorth ), Order( Descending ), Replace Table );
// icing on the cake - append a script so that row selections
// in this table will generate column selections in the source table
Eval(Parse(Eval Insert("\[
candDt << New Script(
"Select Columns",
cdt = Current Data Table();
rows = cdt << Get Selected Rows;
lstNames = Column(cdt,"Term")[rows];
Data Table("^dtName^") << Clear Column Selection;
For (i=1,i<=NItems(lstNames),i++,
Column( Data Table("^dtName^"), lstNames ) << Set Selected(1)
)
)
]\")));
The end result is a table that looks like this:
I included the code for the Select Columns script attached to this table as I find it particularly useful. So for example if you select the first three rows on this candidates table then running the script will select the corresponding columns in the source data table. Useful if you have loads of columns.
- Dave
