*Enlace a Spanish version
Materials: [ Cód.: TLS3x2p4English.mlx ] [ PDF ]
This video is a continuation of the video [
The issue discussed here deals with analyzing what happens when the scaling, made prior to SVD, is incorrect. Unfortunately, the result is a biased model. In fact, the bias is even greater than in the classical least squares model that is often used to justify TLS in many classrooms (even mine)... if the scaling is poorly designed, the cure may be worse than the disease. This may be a problem in PCA/TLS applications.
Since the model is multivariable, and implicit (equation between five variables , the command subspace is used to determine the angle between the subspaces in which the variables can be there if they verify the model’s equations (the null space of ), so the magnitude of the ”bias” is reduced to a single number.
*Link to my [ whole collection] of videos in English. Link to larger [ Colección completa] in Spanish.