Re: degrees of freedom for chi-square testing of goodness of fit
- From: Tim Little <tim@xxxxxxxxxxxxxxxxxxxxxxxxxx>
- Date: Mon, 24 Mar 2008 00:03:50 -0000
On 2008-03-23, frank <frank.degeeter@xxxxxxxxx> wrote:
I start from measured data (all normally distributed)
With equal and known variance?
Accordingly, I calculate exponential fits through these points. I
want to find out which of the six measurements provide the best
result, so I calculate the fit from all possible combinations of
these 6 points.
Just checking my assumptions here: You fit an exponential curve to a
subset of data points. Presumably every subset of 2 or more points,
using least-squares regression?
I would use a chi-square test for that purpose.
The chi-squared test is a hypothesis test. The null hypothesis is
that the calculated statistic is distributed according to a
chi-squared distribution with k degrees of freedom, which can be
derived as a sum of squares of k unit normal variables.
How do you calculate your statistic? In the usual case for using
chi-squared test, under the null hypothesis the data points are
binomially distributed (hence approximately normal) with known mean
and variance. What is your null hypothesis?
- Tim
.
- Follow-Ups:
- References:
- Prev by Date: math : a calculus conjecture
- Next by Date: Re: Manuals
- Previous by thread: degrees of freedom for chi-square testing of goodness of fit
- Next by thread: Re: degrees of freedom for chi-square testing of goodness of fit
- Index(es):
Relevant Pages
|
