nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

White Test for Heteroscedasticity

MikeD_Anderson

Anybody know if we can do this test in JMP or Pro... was asked by a customer. White Test - Wikipedia

I found an article on the SPSS website for a procedure for something similar: Procedure

Thoughts from the team?

Accepted answers

jiancao

Unfortunately JMP Pro doesn't have it or the White heteroscedasticity-consistent standard errors. I'd highly recommend JMP implements both the White test and MacKinnon and White Estimators.

All comments

jiancao

Unfortunately JMP Pro doesn't have it or the White heteroscedasticity-consistent standard errors. I'd highly recommend JMP implements both the White test and MacKinnon and White Estimators.

jiancao

Additional comments: I just found this discussion https://community.jmp.com/t5/Discussions/Test-for-homo-heteroscedacity-in-JMP/m-p/7737#M7731. What he describes is what’s stated in the SPSS article. However, if the test is rejected,which is quite often, then people would want the heteroscedasticity-consistent standard errors reported.

Byron_JMP

this would be a useful test to include when looking at linearity and bias.

In a repeated measures study across the linear range of a test method, often the variation is proportional the the magnitude of the input variable. With this test I think I could answer, "Is the heteroscedasticity through the range significant?"

That would be a good thing to know.

DonMcCormack

When I was at Amgen, I wrote a short script for Breusch-Pagan. If it's of any use, here you go:

tbl = Current Data Table();

cDlog = Column Dialog(
	rCol = Col List("Residual", Max Col(1)),
	pCol = Col List("Predicted Value", Max Col(1))
	
);
residCol = Column(cDlog["rCol"]);
predCol  = Column(cDlog["pCol"]);

tbl << Fit Y by X(
	Y(residCol),
	X(predCol),
	Fit Line({Save Residuals})
);
Column(tbl,N Col(tbl)) << Set Name("BP Resids");
bpResidCol = Column(tbl,N Col(tbl));

n     = N Row(tbl);
zVec  = predCol << Get as matrix;
eVec  = (:BP Resids << Get as matrix);
e2Vec = eVec^2;
iVec  = J(n,1,1);
zVec  = iVec || zVec;

uProd = (e2Vec-Mean(e2Vec)*iVec);

eSqr = eVec`*eVec/n;
outVal = Summation(i=1,N Row(),
	(residCol[i]^2 - eSqr)^2
);
e2Var = outVal/n;
bp    = (uProd`*zVec*Inverse(zVec`*zVec)*zVec`*uProd)/e2Var;
pVal  = 1-Chi Square Distribution(bp,2);
Show(bp,pVal)

Byron_JMP

Don - You rock!

Thx^9,