comparing two bootstrap distributions
wolfzinke_at_gmx.net
Date: 02/02/05
- Previous message: Marc Schwartz: "Re: 3D Surface plot in R"
- Next in thread: Richard Ulrich: "Re: comparing two bootstrap distributions"
- Reply: Richard Ulrich: "Re: comparing two bootstrap distributions"
- Messages sorted by: [ date ] [ thread ]
Date: 2 Feb 2005 03:28:10 -0800
I need some advice in a procedure, where I am not sure it is valid the
way I am using it.
I want to compare the parameter of curves (gaussian) fitted to two data
sets.
Both data sets consists of 8 mesured values, where I have got 20 data
points. Original I did the fitting using weighted least squares to the
mean values for each of the eight values.
To get confidence intervals I used a bootstrap procedure. 100 times I
took a new sample for each value (sampling with replacement)and fitted
the curve to the mean values for the new samples. For each fit I
determined the fitting parameter (amplitude, sigma, center, offset),
thus I ended up with 100 values for the parameter for each of my data
sets, which allows me to determine the median value of each parameter
and its confidence intervals.
Now I want to test each paramter, whether they are significant
diffferent from each other. The obvious method would be to use a
Wilcoxon sign rank test, where I compare the 100 values from each
parameter. Is this a valid method or do I need somehow to consider the
fact, that the distributions, I want to compare, are a result from a
bootstrapping procedure?
I hope someone is able to give me some advice or maybe references for
further information.
Thank you for the support.
- Previous message: Marc Schwartz: "Re: 3D Surface plot in R"
- Next in thread: Richard Ulrich: "Re: comparing two bootstrap distributions"
- Reply: Richard Ulrich: "Re: comparing two bootstrap distributions"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|