Re: How to compare if two fits are significantly different from each other
- From: glenbarnett@xxxxxxxxxxxxx
- Date: 24 Oct 2006 18:38:31 -0700
youngjin.michael@xxxxxxxxx wrote:
Hi,
I am looking for a means to determine how two fits are significantly
different from each other.
I have data and made two fits to my data using different fitting
functions. The question I have is how to determine whether one fit is
significantly better than the other one.
I thought about comparing their Chi sqaure values. This way, I could
say that which one is a better fit, but I could not say how much better
(significance) it is compare to the other one.
Can anyone help me with this problem?
Thanks in advance.
Young-Jin
The big problem with using chi-square statistics is that they're only
comparable if they have the same number of parameters in the fit (so
the DF of the chi-squares are the same). In that case, you can use them
as a measure of fit (though perhaps not the best measure - it depends
on what you're doing).
Nested models (i.e. where one model is a special case of the other) are
usually pretty easy to deal with, and there's lots of information
available about that.
Nonnested models are trickier. There are some statistics designed for
comparing non-nested models.
What two models are you comparing?
Glen
Glen
.
- References:
- [Q] How to compare if two fits are significantly different from each other
- From: youngjin.michael@xxxxxxxxx
- [Q] How to compare if two fits are significantly different from each other
- Prev by Date: Maximum Likelihood estimation
- Next by Date: Re: Maximum Likelihood estimation
- Previous by thread: Re: How to compare if two fits are significantly different from each other
- Next by thread: Unbiased Estimate of Mean for Unequal Probability Sampling
- Index(es):
Relevant Pages
|
