When I use image analyzer add-in to convert the spectral result picture to data table.
https://community.jmp.com/t5/JMP-Add-Ins/Image-Analyzer/ta-p/23886
Because the spectral line is too thick that it always with several Ys on the same X 。
Is there any way to get more accuracy Y from those data?
I draw a normal distribution line as an example, as you can see below, the line pass through X=330 with several Ys, and I use the B parameter as color variable to show B value of each location.
I try take Y as mean of Y, median Y, set B as weight of Y, but still can't get good enough result.
My goal is making the bias (or residual) lower than 0.0025, since the spectral need more accuracy result.
Below is my result so far, the bigger bias around 0.004, I also attached the data for reference.
I wouldn’t use this add-in for this. The artifacts from the pixelization are going to cause you all kinds of trouble. If you want to digitize a plot (spectral or otherwise) you’re going to get better results from the digitizing utility in Brady’s Data Table Tools Add-in (https://community.jmp.com/t5/JMP-Add-Ins/Data-Table-Tools-Add-in/ta-p/28582).
That said, you could try going through graph builder and using the Savitzky-Golay smoother in the spline options. That might do a better job than a bin-wise treatment.
In the screenshot below I did a rough pass for the same approximate region (except mine is plotting upside down from yours, so maybe transmission vs absorbance?) using the SV smoother:
M
