I agree that there are many assumptions either way this is done. One also
has to watch out for biases introduced in the tensor directions...averaging
first does introduce some bias. But this may be mitigated by a decreased
variance. Just my two cents.
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask]] On Behalf Of hongtu Zhu
> Sent: Thursday, January 05, 2006 10:12 AM
> To: [log in to unmask]
> Subject: Re: [SPM] DTI tensor calculation
>
> Never average over repetition and solve 6 linear equations.
> You need a lot of assumptions in order to do this. ...
>
> Quoting dqqiu <[log in to unmask]>:
>
> > Dear SPMers:
> > I'm analyzing a normative dti data set imaged with 6 gradient
> > directions
> > and 4 repetitions. For calculation of the tensor and FA, I wonder
> > whether it is better to average over repetition first and then
> > solve 6
> > linear equations, or just to put all together and solve
> > overdetermined
> > linear equations using Least Square Estimation.
> >
> > I know the question is not really of SPM, but I don't know other
> > place
> > to ask.
> > Any suggestion is appreciated!
> >
> > Thanks in advance.
> >
> > Regards,
> > Kenneth
> > The University of Hong Kong.
> >
> >
>
|