26 feb. 2018

PCR vs. PLS (part 9)

One way to understand the high correlation for the 6th term vs. Ash parameter is to check the 6th term. Terms are spectra which reconstruct the original spectra and are included in the P matrix.
 
The original spectra is treated with first derivative where the peak maximum is converted to a zero crossing, so it is difficult to understand easily what it is going on. So we have to find to which wavelengths correspond the zero crossings.
 
We can check in the loading vector in which wavelengths the values goes from negative to positive or from positive to negative or to use the identify function in R like in this case to find the data points for the zero crossings and to check after to which wavelengths correspond those data points.

identify(wavelength,NIR_princomp$loadings[,6])
[1]    39    80    85    113   132   145   148   
nm    1176  1258  1268  1324  1362  1388  1394  
      
       182   193   200  214   217
nm    1462  1484  1498 1556  153
             Cellulose
It is not the idea of this post a scientific research of course, but we can try to check in the bibliography to have an idea of what this bands are related to in order to understand better our model.
In this case 1484 and 1498 are normally bands assigned to cellulose.


The original spectra treated with the first derivative is:

Remember from the previous post that the ash content is 0 to 10 for blue spectra, 10 to 20 for green, 20 to 30 for orange and 30 to 40 for red, grey for samples with no ash reference value.
In this set we have mix meat meal, pork, poultry,...

No hay comentarios:

Publicar un comentario