Dear all,
I am using the Imodpoly algorithm with python to fit very noisy fluorescence data. However, in a few instances, I notice that changing the polynomial degree or using arpls algorithm will fit my data better. If I am running many data sets and my goal is to perform quantitative analysis and comparison, do I have to use the same fitting algorithm for each data set or can I mix and match algorithms for better fitting?
Thanks
You could make use of the following reference.
Gibson, R. L., Simmons, M. J. H., Stitt, E. H., Horsburgh, L., & Gallen, R. W. (2022). Selection of formal baseline correction methods in thermal analysis. Chemical Engineering & Technology, 45(2), 238–248. https://doi.org/10.1002/ceat.202100120
Good luck,
Hello Sean Wang, as a general practice, before running any training and prediction for any algorithm, all data sets has to be representative to the phenomenon you are looking to presict. Then the training set has to representative to all the different environments your data collected from. Finally if you know that the data sets have some biases that requires corrections to match a known reference, then all corrections has to be performed before using any of these data set for traing and prediction. We always keep in mind that perfect algorithms predict perfectly all the wrong representations that the input data has. Therefore, ensuring that the data is correctly representing what we are looking for is vital before an further analysis. All the best wishes.
Polynomials fit better with everything, but they don't explain anything.
What is the expected relationship between your variables?
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python