627160_en-US

liliconcom

How to pass a variable in the formula of custom profiler

We can define a custom formula in the custom profiler and it is quite easy, as the following example shows

 

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Tiretread.jmp" );
obj = Custom Profiler( Y( :Pred Formula ABRASION, :Pred Formula MODULUS ) );
obj << Objective Formula(
 :Pred Formula ABRASION + .5 * :Pred Formula MODULUS
);
obj << Optimize;

Now, let's say that I would like to change the coefficient value from 0.5 to another setting. So I define the following formula:

 

coef=4;
myFormula=Parse(eval(EvalExpr(":Pred Formula ABRASION + "||Char(coef)||" * :Pred Formula MODULUS")));

the result in the log is ok:

:Pred Formula ABRASION + 4 * :Pred Formula MODULUS

but, if I use the same sentence in the previous statement, I do not succeed to have the expected result:

obj << Objective Formula(Parse(eval(EvalExpr(":Pred Formula ABRASION + "||Char(coef)||" * :Pred Formula MODULUS"))));

I probably did a basic mistake, but I am having trouble with it. and I would like to adapt this method to a more complex situation (change desirability coeffcients in the native expression)
Thank you for your suggestions

Re: How to pass a variable in the formula of custom profiler

Florence,

In general, it's preferable to treat expression as expressions rather than text string to be evaluated. This will work:

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Tiretread.jmp" );
obj = Custom Profiler( Y( :Pred Formula ABRASION, :Pred Formula MODULUS ) );


cVal = 4;

Eval(Substitute(
 Expr(
  obj << Objective Formula(
   :Pred Formula ABRASION + coefVal * :Pred Formula MODULUS
  )
 ),
 Expr(coefVal),cVal
));

obj << Optimize;

Don

Re: How to pass a variable in the formula of custom profiler

Hi Don,

Thank a lot for your answer. 
In this basic example, it works pretty well, as my formula is simple. 
In the real life, I have a more complex situation. I want to be able to set the content of desirabilities that I will catch from a data table. 

y1_target_value=175;
y1_min=150;
y1_max=200;

y2_target=500;
y2_min=450;
y2_max=550;



Eval(Substitute(
 Expr(
  obj << Objective Formula(
   Exp((0.25 * Log(Desirability([100, y1_target, 200], [0.0183, 1, 0.0183], Pred Formula Abrasion)) + 0.25 * Log(Desirability([400, 475, 550], [0.0183, 1, 0.0183], Pred Formula Elongation))) / 0.5)
  )
 ),
 Expr(y1_target),y1_target_value
));

obj << Optimize;

if I apply the substitute in this case, I have an error. 

 

and yes.... I want to change the content of the 6 variables (target and spec for the two Y)!!

 

thank you for your suggestions,

Florence

Re: How to pass a variable in the formula of custom profiler

Florence,

The issue has to do with the way the matrix operator (the open and close square brackets) work. The simplest approach I can think of is to create the matrices beforehand and treat them as a whole, rather than trying to construct them in Substitute. This should work

y1_target_value=175;
y1_min=150;
y1_max=200;

y2_target=500;
y2_min=450;
y2_max=550;

y1Mat = Matrix({y1_min,y1_target_value,y1_max});
y2Mat = Matrix({y2_min,y2_target,y1_max});

Eval(Substitute(
 Expr(
  obj << Objective Formula(
   Exp((0.25 * Log(Desirability(y1Mat, [0.0183, 1, 0.0183], Pred Formula Abrasion)) + 0.25 * Log(Desirability(y2Mat, [0.0183, 1, 0.0183], Pred Formula Elongation))) / 0.5)
  )
 ),
 Expr(y1Mat),y1Mat,
 Expr(y2Mat),y2Mat,
));

obj << Optimize;

 

Re: How to pass a variable in the formula of custom profiler

I feel like I'm missing something here. Why use the Expr/Eval, etc. instead of just putting the formula in?

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

y1_target_value = 175;
y1_min = 150;
y1_max = 200;

y2_target = 500;
y2_min = 450;
y2_max = 550;

y1Mat = Matrix( {y1_min, y1_target_value, y1_max} );
y2Mat = Matrix( {y2_min, y2_target, y1_max} );

obj = Custom Profiler( Y( :Pred Formula ABRASION, :Pred Formula MODULUS ) );

obj << Objective Formula(
 Exp(
  (0.25 * Log( Desirability( y1Mat, [0.0183, 1, 0.0183], :Pred Formula Abrasion ) ) + 0.25 * Log(
   Desirability( y2Mat, [0.0183, 1, 0.0183], :Pred Formula Elongation )
  )) / 0.5
 )
);
 
obj << Optimize;

Re: How to pass a variable in the formula of custom profiler

If you look at the expression in the Formula outline, the values for y1Mat and y2Mat appear unevaluated, i.e., as variables:

Exp((0.25 * Log(Desirability(<FONT color="#FF0000">y1Mat</FONT>, [0.0183, 1, 0.0183], :Pred Formula Abrasion)) + 0.25 * Log(Desirability(<FONT color="#FF0000">y2Mat</FONT>, [0.0183, 1, 0.0183], :Pred Formula Elongation))) / 0.5)

BTW, it looks like I had a few mistakes in the previous code. Here's the corrected version

y1_target_value=175;
y1_min=150;
y1_max=200;

y2_target=500;
y2_min=450;
y2_max=550;

y1Mat = Matrix({y1_min,y1_target_value,y1_max});
y2Mat = Matrix({y2_min,y2_target,y2_max});

Eval(Substitute(
 Expr(
  obj << Objective Formula(
   Exp((0.25 * Log(Desirability(y1Mat, [0.0183, 1, 0.0183], :Pred Formula Abrasion)) + 0.25 * Log(Desirability(y2Mat, [0.0183, 1, 0.0183], :Pred Formula Elongation))) / 0.5)
  )
 ),
 Expr(y1Mat),y1Mat,
 Expr(y2Mat),y2Mat,
));

obj << Optimize;

All this said, I'm not sure this is the way to implement different boundaries for the desirability functions. The formula appears correct, but the objective value is missing.

 

Re: How to pass a variable in the formula of custom profiler

Thank you for the tips! it works

Regarding the formula, it is the one defined by JMP when I launch the custom profiler with the prediction formula generated by Fit Model platform. 

The default is the following:

with this  type of prediction formula:

I will check if I have the same results than on native Model Fitting platform.

Thanks again!

 

Re: How to pass a variable in the formula of custom profiler

Thanks Florence,

I found the problem. I used ABRASION and MODULUS in Custom Profiler but used ABRASION and ELONGATION in the Objective Formula.

Here is the (hopefully) error free code:

obj = Custom Profiler( Y( :Pred Formula ABRASION, :Pred Formula MODULUS ) );

y1_target_value=175;
y1_min=150;
y1_max=200;

y2_target=500;
y2_min=450;
y2_max=550;

y1Mat = Matrix({y1_min,y1_target_value,y1_max});
y2Mat = Matrix({y2_min,y2_target,y2_max});

Eval(Substitute(
 Expr(
  obj << Objective Formula(
   Exp((0.25 * Log(Desirability(y1Mat, [0.0183, 1, 0.0183], :Pred Formula Abrasion)) + 0.25 * Log(Desirability(y2Mat, [0.0183, 1, 0.0183], :Pred Formula Modulus))) / 0.5)
  )
 ),
 Expr(y1Mat),y1Mat,
 Expr(y2Mat),y2Mat,
));

obj << Optimize;