282648_en-US

liliconcom

I want to know the result when fitting an specified regression line.

Hi all,

Please let me share your ideas.

Thanks in advance

~~~

I want to know the result when fitting an specified regression line.

Method 1 : fit special (in fit y by x)

There is no R-Sq and the values of Model are '0'. I don't know well the reason why.

 

Method 2 : using teaching scripts

I want to know that there is some way to check my fitting line(equation) and the result here.

 

3. Is there any other methods to check analysis result when fitting an specified regression equation ?

Thanks

Ikju

 

 

Re: I want to know the result when fitting an specified regression line.

I have the sense that there is more to your question than meets the eye. But if you care to resort to JSL, you can do more or less anything. But you probably knew that already! So the 'best' answer will depend on exactly what you are looking for.

NamesDefaultToHere(1);

// Make some x and y data values . . .
n = 10;
dt = NewTable("Data",
  NewColumn("x", Numeric, Continuous, Values(Transpose(0::(n-1)))),
  NewColumn("y", Numeric, Continuous, Formula(10 + 3 * :x + 2 * :x^2 + RandomNormal(0, 1)))
  );

// Define any function of :x you like . . .
f = Expr(10 + 2*:x + 3*:x^3);

// Add the values of the specified function . . .
addFunctionValues = Expr(dt << NewColumn("Function Values", Numeric, Continuous, Formula(fTBD)));
SubstituteInto(addFunctionValues, Expr(fTBD), NameExpr(f));
addFunctionValues;

// Get the residuals and continue on as required . . .
dt = NewColumn("Residual", Numeric, Continuous, Formula(:y - :Function Values));

Re: I want to know the result when fitting an specified regression line.

Thank you. Ian, I will  refer to your ideas. 

Re: I want to know the result when fitting an specified regression line.

Hi @IkjuShin ,

 

What is the goal?  If you specify the slope and intercept in Fit Y by X, then the remaining data can only tell you information about error (hence the DF of 40). Also, there is no R2 ( you already defined the slope) because you specified what the line should look like so there is no "fit" to the data. Are you looking for how well the data fit the specified line?  Are far away each point is from the expected?

 

Thanks,

Re: I want to know the result when fitting an specified regression line.

Hi Chris, 

They are looking for how well the data fit the specified line.

Their data are attached.

 

Re: I want to know the result when fitting an specified regression line.

There is no R-Sq and the values of Model are '0'. I don't know well the reason why.

R-sq must be [0,1]. However, with a user specified intercept, a slope or both, R-Sq may be >1. Thus, JMP doesn't calculate it whenever you've constrained a coefficient.  

 

 

 

 

Re: I want to know the result when fitting an specified regression line.

You can just compare the error SS from the special fit to that from the unconstrained fit. 

Re: I want to know the result when fitting an specified regression line.

Hi Jianchao,

I learned that R-sq could > 1 in some cases like constrained coefficient.  I'd like to check more details through tech support.

Thanks

Re: I want to know the result when fitting an specified regression line.

By the definition provided by Jian, the calculated R-sq for your constrained model is greater than 1. Take a look at the attached file to see. I've also included the standard least fit model to show the R-sq for that model as well. JMP doesn't produce R-sq when the model is constrained.

Re: I want to know the result when fitting an specified regression line.

Hi Duane,

I checked and understood related things.

Many thanks