Prev: Trimesh contour plot with nans
Next: detecting ships at horizon
From: Mitchell Schull on 21 May 2010 13:29
Hopefully I can explain my problem correctly. I am trying to find a way to conduct a joint hypothesis test in matlab. For example I want to test restrictions that equalize all regression coefficients of alpha, and then beta and then gamma of the following regression

y(i,l) = alpha(l)*dummy(l) +beta(i,l)*A(i,l)+gamma(i,l)*B(i,l)

where i is a plot and l is a site.

I have looked all over and it seems no one has done this.

thanks
From: Tom Lane on 21 May 2010 14:51
> Hopefully I can explain my problem correctly. I am trying to find a way
> to conduct a joint hypothesis test in matlab. For example I want to test
> restrictions that equalize all regression coefficients of alpha, and then
> beta and then gamma of the following regression
>
> y(i,l) = alpha(l)*dummy(l) +beta(i,l)*A(i,l)+gamma(i,l)*B(i,l)
>
> where i is a plot and l is a site.
>
> I have looked all over and it seems no one has done this.

A couple of things to look into if you have the Statistics Toolbox:

1. The linhyptest function can perform a hypothesis test involving multiple
constraints on the coefficients. For example, depending on how you code your
dummy variables, you could test that they are all equal or all zero.

2. For models you fit using one of the anova functions, the multcompare
function might be useful.

-- Tom


 | 
Pages: 1
Prev: Trimesh contour plot with nans
Next: detecting ships at horizon