38888_en-US

liliconcom

Confidence and prediction intervals for orthogonal regression?

I can tell that this question was posted back in 2012, but never answered, so here it goes again:

Is there a way to plot confidence and prediction intervals for orthogonal regression in the Bivariate platform. Also I would like to save the interval equations. Both options exist for "Fit line", i.e. OLS linear regression.

If the option does not exist, I would like to request it be implementet. 

Re: Confidence and prediction intervals for orthogonal regression?

Hi Ian,

 

Thank you for this JSL script. However, the script only boostraps estimates of slope and intercept. I was curious how would I then go about estimating prediction interval in deming regression using this output? I would appreciate any assistance.

 

Thank you

 

 

Re: Confidence and prediction intervals for orthogonal regression?Alas, I don't have Pro :'-(Re: Confidence and prediction intervals for orthogonal regression?

With JMP Pro, you could try bootstrapping:

NamesDefaultToHere(1);

// Sample data
dt = NewTable("Orthogonal Regression",
  NewColumn("x1", Formula(RandomUniform(-1, 1))),
  NewColumn("x2", Formula(RandomUniform(-1, 1))),
  AddRows(50)
);

// Fit orthogonal
fo = dt << Bivariate(Y( :x2 ), X( :x1 ), Fit Orthogonal( Equal Variances ));

// Bootstrap fitted parameters
dt2 = Report(fo)[TableBox(2)] << Bootstrap(1000);

// Look at the result
dt2 << Distribution(
    Continuous Distribution( Column( :Intercept ) ),
    Continuous Distribution( Column( :Slope ) ),
    SendToReport(
     Dispatch( {"Intercept"}, "Quantiles", OutlineBox, {Close( 1 )} ),
     Dispatch( {"Intercept"}, "Summary Statistics", OutlineBox, {Close( 1 )} ),
     Dispatch( {"Slope"}, "Quantiles", OutlineBox, {Close( 1 )} ),
     Dispatch( {"Slope"}, "Summary Statistics", OutlineBox, {Close( 1 )} )
    )
   );

Re: Confidence and prediction intervals for orthogonal regression?

Noone from JMP reading this?