Error message with confidence interval calculation

[Siva]

Hello Everyone,

I am quite new to SPINEVOLUTION. I get the following error message
while calculating confidence interval on the fit parameters.
--------------------------------------------------------------------------------------------------------------------------------------------------
Error: in order to save time, do not compute the covariance matrix or
the confidence intervals when using the shaking technique or
performing full optimization for multiple starting points.
--------------------------------------------------------------------------------------------------------------------------------------------------

I am using the default NL2SOL method.

How do I fix this?

I have another question. Is there a way to specify chemical shift
parameters in "ppm" units rather than "kHz" in the fit_par list?

Thanks

Sivakumar

[Mikhail Veshtort]

Hi Sivakumar,

The error message you are getting is self-explanatory. If you need to
compute the covariance matrix in the situations stated in that
message, you should compute it afterwards, when you already computed
the optimal fit parameters. Just use these parameters as the starting
point and request covariance matrix calculation. If you don't need the
matrix, then just don't ask the program to compute it.

In order to fit chemical shifts in ppm, fit them as user-defined
variables. For example, let CS1ppm be you chemical shift in ppm on
spin 1. Specify it as a fit parameter, and compute cs_iso_1 from in
the Variables section.

Mikhail

[Siva]

Hi Mikhail,

Thanks for your quick response. But it is not clear to me what you
mean by computing the confidence (or covariance) "afterwards". I tried
the following things. First I did the simulation with fit_par and then
redid the simulation with by removing the fit_par statement and used
the best-fit parameters reported in the ".par" file and also included
the "-confint" in the options line. It did not produce any confidence
intervals.

Could you please tell me how to do it correctly?

Thanks

Sivakumar

[Mikhail Veshtort]

Hi Sivakumar,

-confint option obviously works only when you are optimizing/fitting
something. So, if you have not specified any fit parameters, the
option will be simply ignored. In your case, you have to use it with
your best-fit parameter values (from the first step) specified as the
initial point of a fit. But, at this step, you should not do any
shaking (-shake option) or multiple starting points, or you will get
an error message as before. If you specify a grid of starting points
for parameter optimization (by saying e.g. fit_par x/0:10/), you
should disable this grid when you are computing the covariance matrix/
confidence limits in the second step (by replacing that line with
fit_par x).

Mikhal

[Siva]

Thanks, Mikhail!

It works fine.

Sivakumar

[Antonio.N...@roma1.infn.it]

Hi,
could anybody please explain to me why I am getting the message that the 3.4
version is expiring on the 1st of september 2009? Thanks a lot.
Regards
Antonio

----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

[Mikhail Veshtort]

Hi Antonio,
I just forgot to change this from the previous version... I also have
found a couple more small bugs in the new version, and I will soon
release an update (together with the versions for the remaining
platforms, etc). The actual expiration date is September 2010.
Mikhail