Hello,
I'm not sure what is going wrong here. It looks to me, from the code,
that Analysis is taking account of the sign, but from what you are saying
it must not be somehow. If you could send me (off list) a sample few
planes of spectra (so a pseudo-3D spectrum or a few 2D spectra, depending
how it was recorded) with negative values then I can take a look.
On the issue of allowing the user to input suggested initial values, I
have considered that but haven't implemented that in v2 (yet!). I was
definitely going to allow that in v3. But it would be better if the
fitting worked without that (because it would be a pain if you had a
reasonable number of peak groups).
Wayne
On Fri, 10 May 2013, Christoph Hartlmüller wrote:
> Hi CCPN users,
>
> I have some trouble using the Rate Analysis function in Analysis (Version
> 2.2. Release 1). I managed to setup an NMR series, fit a function and look at
> the plots of the fits. The problem concerns the fitting:
> I have a series of HSQC spectra that were recorded using different delay
> times. In this case, we observe signals that increase with increasing delay
> times. So the data that we want to fit is of the type (I selected this type
> in Analysis)
> A*exp(-Bx) + C
> where B and C are positive, A is negative and the absolute value of C is
> roughly equal to that of A
> Now the problem is, that in the fit results Analysis only gets positive
> values for A and negative values for B and C. So the fitting curve is
> completely wrong.
> So the questions are basically: Is there a possibility to set starting values
> for the parameters A,B and C or limited them to a certain range? If no, is it
> possible to setup a custom equation, like -A*exp(-Bx) + C ?
>
> To narrow down the problem:
> If I choose some other equation (for example linear) I can see that Analysis
> roughly gets the data fitted (as much as possible of course). So the setup is
> ok, it's just that the fitting procedure is not getting anywhere near the
> correct solution.
>
> Also, I phase-shifted the spectra 180 degree to make everything negative,
> since A and C should then switch sign. However the fitting does not work any
> longer since the result list is complete empty. My guess is that Analysis
> does not pick the peaks in the other HSQCs anymore as they are now negative (
> I did switch "Peak=> Peak Finding => Extrema to search for" to "positive and
> negative" ; that's how I created the peak list at the first place )
>
>
> Thank you and best regards,
> Christoph Hartlmueller
>
|