LabVIEW
How do I fit a curve with two independant variables
Well, what is your definition of a 2D polynominal? Do you have an actual model function for your data?
(e.g.: z = Ax + Bx^2 + Dy + Ey^2 + Fxy + G ...)
I don't think this is directly available in LabVIEW, but it would not be too difficult to adapt the "nonliner Lev-Mar Fit.vi" for this use. I use it for all kinds of much more complex functions that that. 😉
In fact I did exactly the above described function last weekend and made a very simple demo using Lev-Mar subVIs that I modified from the stock versions a few years ago to allow fitting of arrays. In the case of arbitrary x-y data that is not on a regular grid, you would need to use extra connectors for the independent variables, etc.
(I have much more elaborate modified versions of
Lev-Mar for my own use, but the attached should be OK to show the principle.)
Enjoy!
Let me know if you have questions. There are probably a few minor errors in the code, but it seems to work just fine. Simply unzip the llb and run it (LabVIEW 7.1).
(e.g.: z = Ax + Bx^2 + Dy + Ey^2 + Fxy + G ...)
I don't think this is directly available in LabVIEW, but it would not be too difficult to adapt the "nonliner Lev-Mar Fit.vi" for this use. I use it for all kinds of much more complex functions that that. 😉
In fact I did exactly the above described function last weekend and made a very simple demo using Lev-Mar subVIs that I modified from the stock versions a few years ago to allow fitting of arrays. In the case of arbitrary x-y data that is not on a regular grid, you would need to use extra connectors for the independent variables, etc.
(I have much more elaborate modified versions of
Lev-Mar for my own use, but the attached should be OK to show the principle.)
Enjoy!
Let me know if you have questions. There are probably a few minor errors in the code, but it seems to work just fine. Simply unzip the llb and run it (LabVIEW 7.1).
Here is the LabVIEW 7.0 version.
Just glancing at it, your function does not look like a simple polynominal, but is quite a bit more complex. (Unless it can be rewritten in some other form, which I haven't tried). 🙂 Note that my algorithm should have no problems dealing with your function. Do your function values form a regular grid or are they randomly distributed?
(Just a note to the demo: The function I quoted (z=Sum(A(n)x^py^q)) above could be easily implemented using the "general LS fit", which is a direct method. You would just index all your (x,y) pairs, then generate the "H" matrix by creating an array where each column (n) is the function with all A except A(n) set to zero and each row is the point index. If course thi
s will not work if the function is not a "k-dimension linear curve", as is the case with most of my problems. Lev-Mar converges in a few steps, so it's probably not worth re-inventing the wheel.)
Just glancing at it, your function does not look like a simple polynominal, but is quite a bit more complex. (Unless it can be rewritten in some other form, which I haven't tried). 🙂 Note that my algorithm should have no problems dealing with your function. Do your function values form a regular grid or are they randomly distributed?
(Just a note to the demo: The function I quoted (z=Sum(A(n)x^py^q)) above could be easily implemented using the "general LS fit", which is a direct method. You would just index all your (x,y) pairs, then generate the "H" matrix by creating an array where each column (n) is the function with all A except A(n) set to zero and each row is the point index. If course thi
s will not work if the function is not a "k-dimension linear curve", as is the case with most of my problems. Lev-Mar converges in a few steps, so it's probably not worth re-inventing the wheel.)
Micah,
I had quite a bit of trouble converting it back to 6.1 and it lost some functionality. When I first opened it in 6.1, I got an insane error and a few things were broken. I had to delete and redo the event assignements and it is now working.
Since 6.1 does not have signaling properties, you need to manually move one of the sliders to update the graphs after starting the program (I did not bother to code around this limitation). The rest seems to work fine.
(Whenever I go back to 6.1 or earlier, it amazes me how much easier, more functional, and better the 7.1 version is. Try to upgrade if you can.)
I had quite a bit of trouble converting it back to 6.1 and it lost some functionality. When I first opened it in 6.1, I got an insane error and a few things were broken. I had to delete and redo the event assignements and it is now working.
Since 6.1 does not have signaling properties, you need to manually move one of the sliders to update the graphs after starting the program (I did not bother to code around this limitation). The rest seems to work fine.
(Whenever I go back to 6.1 or earlier, it amazes me how much easier, more functional, and better the 7.1 version is. Try to upgrade if you can.)
