I have a customer who has collected coordinates of a device at various insertion distances and would like to fit a model / models to the data in order to obtain estimates of where the device will be. Is this something that anyone has experience with?
You could try partial least squares, with the X,Y,Z coordinates as the response and the insertion depth as the factor.
Also, you could treat the X,Y,Z coordinate points as a curve and use FDE with the insertion depth as a supplementary variable.
Perhaps a simpler approach would be to model the coordinates X, Y, Z independently as a function of insertion depth. You lose the correlation dependencies that the X, Y, and Z have on each other, but that may not matter.
Thanks, Sam.
With your FDE idea...how would X, Y, Z be the Y and X inputs (shape)?
