Spiky spectra vs PCA reconstructed:

here is some examples of the spiky spectra with very low BH mass estimates in which we reconstruct the spectra and then we have a normal range BH mass. here the relative difference between after and before PCA applied BH mass is more than 70%.

eta=(M_BH(ab)-M_BH(bb))/M_BH(ab) > 0.7  assuming that M_BH(ab) is correct.

there are cases where eta is larger than 0.85 or 0.9 but the row data has significant problem (missing data around 2800 Angs or absolute noise)

black curve is the row data

red curve is the reconstructed spectra

the variable "vari" in the left panel is the relative differences between masses.

 

 

and these are samples for the case when there is very low differences between before of after PCA applied: the relative differences between BH mass are smaller than 10%