2 ms·
Oh dear! He does not even discuss the HUGE errors that may appear when taking logarithms... If you have points near 0 and away from 0 the former get an artifici
by pfortuny 12y ago
Oh dear! He does not even discuss the HUGE errors that may appear when taking logarithms... If you have points near 0 and away from 0 the former get an artificial enormous weight and bias the "interpolation" too much towards them.
CAREFUL!
Basically: the mean value of the logarithms is not the logarithm of the mean value.
- rprospero 12y agoWouldn't the arithmetic mean of the logarithms be the logarithm of the geometric mean? 1/n Σ ln xᵢ = 1/n ln (Π xᵢ) = ln ( n√(Π xᵢ) )