Is total least squares always better than ordinary least squares?
Not necessarily. Total least squares is most beneficial when significant errors are present in *all* measured variables. If the errors are primarily concentrated in the dependent variables, ordinary least squares might be sufficient and computationally simpler. The choice depends on a careful assessment of the error structure of the data. Overusing TLS when not needed can introduce unnecessary complexity and computational burden.
Ask Sabine Van Huffel the follow-up →