Pharmacokinetics Dose Response

[ptc-1368288]

Under proofing, will be saved in several versions,
expected within a day or two. Wait for questions.

Walter, are you still with the collab ?

jmG

[TheodoreM.Bones]

Amid all this thread, I would plaintively like to ask what is the utter use of trying to fit some raw data into a DE template if you have already fitted the data with some model to a tight fit? A least squares fit does not have the vagrancies of a Minerr fit with chosen guess values. Using MathSoft�s Quick Sheets on Variation of Parameters method, I have tried fitting a simple cubic, or even a quadratic, with a second order DE, and got no fixed solutions. The DE fit was good only at the starting point. If you have a tight SSE fit, you can find the derivatives from it easily, but no need of them.

A Minerr fit is like a smogsboard. A bit of everything in the program.

Besides, these Runge-Kutta, etc. methods require evenly spaced data. How are you going to get raw data like that?

[ptc-1368288]

>Amid all this thread, I would plaintively like to ask what is the utter use of trying to fit some raw data into a DE template if you have already fitted the data with some model to a tight fit ? <<br> _________________________

That's exactly what it is, once the function is found there is no need for the passage via all that monkey business of DE that is not right most probably. Yes, a long exercise that nearly exhausted my patience but so rewarding to offer the "Dose Response", and there is damned nothing in the web except for that very introduction from Wikipedia that does match in the work sheet.

I understand that not all collabs will catch it by first reading. I can explain and entertain for pages, however it addresses to a very limited audience. The size of the data (13) is finally plenty and the accuracy is not detrimental at all. That last point should recomfort Walter especially about so much statistics applied.

The attached is saved in 8, it should work as it does not involve numerical solvers. However, if the Minerr of the user version does not work as per 11 (created), then for a new data set ( a new Dose Response) it should be adapted.

Jim, please let me know if it works for you.

Oh ! a more "Newphi proof" procedure could be added, but this function seems so specific to the medical domain.

jmG

[ptc-1214470]

THe worksheet works well except for lacking the definition of scale(M,Min,Max).
I added the definition (see attachment) and it works fine,
Thanks, Jean
Jim S.

[ptc-1368288]

On 7/8/2009 10:08:09 AM, Offroc wrote:
>THe worksheet works well
>except for lacking the
>definition of
>scale(M,Min,Max).
>I added the definition (see
>attachment) and it works fine,
>Thanks, Jean
>Jim S.
_______________________________

Right Jim,

scale (M,Min,Mx) is built-in from the Improc,
same as the Scale(M,Min,Max) built-in plain Mathcad.
My negligence.
Glad it works fine.

Jean

[TheodoreM.Bones]

Changing the values in the XY file slightly, then running the program, the result was exactly the same. No change.

This is a canned soup program. What it says on the label is what you get. I would rather eat in the kitchen and help myself to what is on the burner.

[ptc-1368288]

>Changing the values in the XY file slightly, then running the program, the result was exactly the same. No change.<<br> _____________________

That's right, a unique property of this function, i.e: if the data set delimitates or border the function, the function is unique and the accuracy of the data does not matter much. The other point unclear is about the spacing of the data and the population. This function would need a complete piece of maths to front end. My interest now is about how to initialise wisely. Got some basic ideas, that need be confirmed. By appropriate initialising, the function and it's recovery will simplify greatly the project vs experimental data of various length and amplitude. OK, we are both newbies in there, this kind of function has never been asked for a fit in the collab.

jmG

[TheodoreM.Bones]

I must have forgotten. Here are 2 fits to that raw data.

[ptc-1368288]

On 7/8/2009 6:12:44 PM, bones7xx wrote:
>I must have forgotten. Here
>are 2 fits to that raw data.
_____________________________

The matter is not to fit a data set, rather to retrieve the function and calibrate the function on at least one (1) reference point, like calibrating an in-line instrument. For my personal interest in curve fitting, I will devote some time to exercise about initializing such a recalcitrant function vs what the Mathcad maths can do.

Thanks for your interest Theodore.

Jean

[ptc-1368288]

Just done and revamped accordingly to the auto-guess of the initial parameters and tutored. The PWMinerred coeff's are left manual plug-in. This project, I certify for public use. Mona, if you catch it, please feel free to add to the PTC library.

Truly yours to all readers

jmG

[TheodoreM.Bones]

I am starting a new thread on the problem submitted by Walter of Graz University for the GFR or renal clearance of the kidney as determined with a non-metabolicized marker, which passes out with no re-absorption in a zero-comportment model. The essential task was to find the half-life time of clearance. Later, experimenting with the rkfixed MathCad algorithm, it was found that a very complex first order ODE was necessary for a solution and it was, therefore, useless to get a symbolic solution with an assumed ODE.
Much re-ordering of the data was needed in this work, but the relative errors were about 2% and thus the work was acceptable. The assumption that the half-life was � of the initial value is proven wrong - my earlier mistake.

This thread has no application to dose response time. In that process the drug being tested is a drug that accumulates in the blood stream and has only an approximate half-life, depending on diet and other factors.