Peak integration improvements

[zeehio]

• In the tutorial, repeat peak detection after alignment and normalization and before integration. It improves results because peaks shift on alignment.

• In the tutorial and docs, explain different strategies for peak integration and consequences.

  • Using the peaks from just one sample is a bad idea, unless that sample contains all possible peaks (pool of samples). In the case is a pool this may still not be the best approach.
  • Integrate peaks from all samples and merge peaks. We have room for improvement here in the pkg

Laura did the lorentzian peak fitting and it's already in the package, we are using it. I want to do some plots to show its performance. Let's see some plots of the citrate region described in the original paper from the dataset of the tutorial and compare it to ours, besides the BCAA. I'm thinking show the spectra, show the peak detection, show the lorentzian fitting, show the peak areas, show missingness in detected peaks

• Show a plot with peak integrations

Now that the peak detection fits lorentzian curves and estimates areas, we should use those debug plots to show how well they work.

• Improve peak width estimation. Currently we base the estimation on the alanine peak, but we use the FWHM. I would probably 3/2 the FWHM so we integrate 90% of the peak, not just the top half (re-check the 3/2 based on a lorentzian distribution)

Again let's check the area estimation from Laura first

• Use wavelets for peak width estimation?

Checked as not done, let's show how Laura's lorentzian fitting works it's most likely good enough and therefore we dont need to bring in wavelets