With the "Corrgram" package we can see patterns that can help us to recognize possible inter-correlations in a big matrix. This could be the case to see the correlation to every wavelength respect to all others. This way we can see the high correlation respect to their neighbors and between the different overtones.
I have selected just 40 (in steps of 2nm) consecutives wavelengths in this case.
Interesting, but I think a heatmap is easy to read & allows much higher data density - one pixel per wavelength, if necessary.
ResponderEliminar- What advantages do you think corrgram has compared to this solution?
- Even with corrgram: Why not use the ellipses to show the extent and direction of correlation?
Thanks for the comment. I´m quite new in R, and as suggested in another comment by Kevin W. I tried the option Corrgram. But I´ll check what you suggest as well.
EliminarYou are rigth is better with ellipses.
Hi, I am frequently use corrgram and wonder how we could build the correlation matrix with missing data. Thank you!
ResponderEliminarThanks to you.
EliminarIn the case of wavelengths I had no problem because there are not missing data. But with the constituents I had to delete samples with missing data.