- #1
stoked
- 1
- 0
Hi,
I'm posting this in this particular forum because, though this's a statistics question, my application is in high energy.
My question is regarding a problem in Bevington's book (Data Reduction and Error Analysis..., Page 193, Ex. 10.1), but I'll give a general description here...
Say you want to fit to a scattering distribution function: f = (a + b cos2 theta) using the likelihood method. cos[tex]\theta[/tex] ranging from -1 to 1, you would get a normalization integral, norm = 2(a + b/3). However, what goes into the actual code is the normalized fuction (a + b cos2 theta) / 2(a + b/3), so that the fitter is only sensitive to (a/b) and not a and b separately. So if you were asked to estimate a and b (as does Bevington's problem)...what do you do?
I am aware of the "extended" likelihood method. Is that applicable here? If so, how?
Thanks a bunch.
I'm posting this in this particular forum because, though this's a statistics question, my application is in high energy.
My question is regarding a problem in Bevington's book (Data Reduction and Error Analysis..., Page 193, Ex. 10.1), but I'll give a general description here...
Say you want to fit to a scattering distribution function: f = (a + b cos2 theta) using the likelihood method. cos[tex]\theta[/tex] ranging from -1 to 1, you would get a normalization integral, norm = 2(a + b/3). However, what goes into the actual code is the normalized fuction (a + b cos2 theta) / 2(a + b/3), so that the fitter is only sensitive to (a/b) and not a and b separately. So if you were asked to estimate a and b (as does Bevington's problem)...what do you do?
I am aware of the "extended" likelihood method. Is that applicable here? If so, how?
Thanks a bunch.