[Date Index] [Thread Index] [Author Index]

Re: Linear regression, error on both vaules x and y

• To: mathgroup at smc.vnet.net
• Subject: [mg70114] Re: Linear regression, error on both vaules x and y
• From: Ralph Kube <ralph.kube at rwth-aachen.de>
• Date: Wed, 4 Oct 2006 05:59:05 -0400 (EDT)
• References: <efldqq$d9l$1@smc.vnet.net> <efq62a$2ep$1@smc.vnet.net>

Ray Koopman wrote:
> Ralph Kube wrote:
>> Hi,
>> I measured the hall-current of a Ge-semiconductor. The experiment
>> yielded data for the current and for the voltage, both with their
>> own error. Both measurements are uncorrelated.
>> Now both the voltage and the current have individual uncertainties
>> due to the method of measurement.
>> How can I perform a linear fit of this dataset and incorporate both errors?
>> Ralph Kube
> 
> Do you mean the voltage uncertainty varies from measure to measure,
> and similarly for the current uncertainty? Are the uncertainties
> absolute (i.e., standard deviations) or relative (i.e., coefficients
> of variation)? Are they known, or only estimated?
> 

Thank you very much for your answers. I solved the problem
by using Sqrt[ sigma_x^2 + sigma_y^2] as my weights.
It worked well enough.
Ralph Kube