topic I think I see a bug in saving CI border formula in Bivariate: 90% CI saved regardless of alpha chosen. 17.0.0 in Discussions

I think I see a bug in saving CI border formula in Bivariate: 90% CI saved regardless of alpha chosen. 17.0.0

l_yampolsky — Mon, 31 Jul 2023 17:24:59 GMT

I was trying to save datapoints that lay above 99.9% CI from Bivariate linear regression results. Selected alpha 0.001 and saved Indiv CIs formula. JM surely enough saves 2 columns named Lower.. and Upper 99.9% Indiv[my Y variable name]. However, when selecting points in which my Y variable > Upper I see that these limits correspond to much less strict CI, apparently 90%.

Removing Fit and redoing it does not help, obviously

Bivariate(
	Y( :"Dtranscr=sqrt(Sum(dZ2))"n ),
	X( :Meanpident ),
	Fit Line(
		{Confid Curves Indiv( 1 ), Line Color( "Red" )},
		{Set α Level( 0.001 )}
	),
	SendToReport(
		Dispatch(
			{},
			"Bivar Plot",
			FrameBox,
			{Frame Size( 617, 465 ), Grid Line Order( 3 ), Reference Line Order( 4 )
			}
		)
	)
);

, because alpha is set after the fitting is done.

This looks like a bug worth fixing, unless I am overlooking something.

Re: I think I see a bug in saving CI border formula in Bivariate: 90% CI saved regardless of alpha chosen. 17.0.0

SDF1 — Wed, 02 Aug 2023 14:48:51 GMT

Hi @l_yampolsky ,

Not sure why you're having an issue. I tried it with the Big Class.jmp file, and it works just fine.

I first did a fit y by x of height and weight, fit a line to it, then changed alpha, then saved the confidence indiv limit formula, and did that for both alpha = 0.001 and 0.01, and it worked just fine. The column formulas are different, and the smaller alpha has a wider CI, as expected.

Are you sure you did the order of operations correctly so that when you saved the Confidence Indiv Limit Formula it saved the correct formula?

Not sure what's going wrong, but give it a try with the Big Class.jmp file first and make sure that what you're expecting happens (no bug) or something different (bug) happens.

Hope this helps!,