relax SVN archive

Got the boot-strapping method to work in test_task_model_par_est_tsmfk01.

The problem was, that simulated values, was extracted from the initial data structure.
Thereby creating an array with similiar values, and a std of 0.0.

The result for 50 boot strap.

param: r2a_err, with err: 0.48131663, compared to MC: 0.49354342, with 50 boot 0.48171919
param: dw_err, with err: 0.30874093, compared to MC: 0.32813703, with 50 boot 0.37605409
param: k_AB_err, with err: 0.64606994, compared to MC: 0.65384783, with 50 boot 0.63808023

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Modified module which estimate model parameter errors, to scale to units in relax.

The R2eff errors come in rad/s. The Jacobian come in rad/s.
The co-variance matrix come in (rad/s)^2. The standard deviation of the parameters the comes in rad/s.

But dw is stored as ppm, therefore a scaling is necessary, and is performed with the scaling matrix from the target function.

Now there is excellent agreement.

Systemtest:
relax -s Relax_disp.test_task_model_par_est_tsmfk01 -d

param: r2a_err, with err: 0.48131663, compared to MC: 0.49354342, with 2 boot 0.00000000
param: dw_err, with err: 0.30874093, compared to MC: 0.32813703, with 2 boot 0.00000000
param: k_AB_err, with err: 0.64606994, compared to MC: 0.65384783, with 2 boot 0.00000000

Juhuuu!

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

To target function, added function which will calculate the Jacobian and its corresponding unit conversion matrix and return them.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Added comments to co-variance module, for explanation of data dimensionality.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

In the lib function of TSMFK01, returned the Jacobian as list of derivatives. This is for easier manipulation in target function.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Fix for the minfx version checking logic in the dep_check module.

Now newer versions of minfx will be handled.

Possible fix for the weight, but as the error is 1, then it does not matter.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Tried to implement bootstrap method.

But that goes horrible wrong.

The numbers are now:
param: r2a_err, with err: 0.48131663, compared to MC: 0.49354342, with 600 boot 0.00000000
param: dw_err, with err: 117.99734537, compared to MC: 0.32813703, with 600 boot 0.00000000
param: k_AB_err, with err: 0.64606994, compared to MC: 0.65384783, with 600 boot 0.00000000

So, something is right, and something is really wrong.

It is clear, that the reported error estimation scales with the error.

So, either one should start with simulations from peak intensity and then forward.
Or, maybe the use of errors are wrong.

There are some rules for how errors can be added together.
If they are a product, then the fractional errors can be summed together.

But, I have to look into this.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Added new data with 5% error, so errors actually correspond to the reported errors.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.

Modified data generator script, to have 5% error, and add the error to the data.

task #7824(https://gna.org/task/index.php?7824): Model parameter ERROR estimation from Jacobian and Co-variance matrix of dispersion models.