Hi Damien @dp30,
What I do for mixture Space-Filling designs (with few factors like 3) to see homogeneity of points distribution in experimental space is to use the Graph Builder and combine several informations on one graph, using 2 factors at the same time (X1 and X2 here) :
Using a dummy column you can plot histogram (kernel density style) next to a contour plot for X1 & X2.
I didn't spend time customizing the colors, but here is the script so you can try it :
Graph Builder(
Size( 711, 588 ),
Show Control Panel( 0 ),
Variables( X( :X1 ), X( :Area ), Y( :Area ), Y( :X2 ) ),
Relative Sizes( "X", [558 99] ),
Relative Sizes( "Y", [112 411] ),
Elements(
Position( 1, 1 ),
Histogram(
X,
Y,
Legend( 19 ),
Overlap( 0.6394 ),
Histogram Style( "Kernel Density" ),
Smoothness( -0.5229 )
)
),
Elements(
Position( 1, 2 ),
Points( X, Y, Legend( 3 ) ),
Contour( X, Y, Legend( 5 ), Number of Levels( 4 ) )
),
Elements( Position( 2, 1 ), Points( X, Y, Legend( 12 ) ) ),
Elements(
Position( 2, 2 ),
Histogram(
X,
Y,
Legend( 18 ),
Overlap( 1.089 ),
Histogram Style( "Kernel Density" ),
Smoothness( -0.4128 )
)
),
SendToReport(
Dispatch(
{},
"graph title",
TextEditBox,
{Set Text( "Visualisation X1 & X2 experimental space" )}
),
Dispatch( {}, "Y title", TextEditBox, {Set Text( "" )} ),
Dispatch(
{},
"400",
LegendBox,
{Legend Position( {19, [4], 3, [0], 5, [1], 12, [2], 18, [3]} )}
)
)
);
I can reproduce it on a toy sample dataset if you wnt to see in details how it is constructed with Graph Builder.
Best,
Victor GUILLER
L'Oréal Data & Analytics
"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)