We've been using the term "normalize" to refer to both the kernel normalization -- denominator = sqrt(k(t1, t1) * k(t2, t2)) -- and to rescaling branch lengths (norm.mode). From now on, "normalize" should be used exclusively in reference to the kernel distance, and the latter should be "rescaling" branch lengths.
Either operation should not affect other distance measures. For example, it should be possible to have a distance function with two or more kernel distances, where one kernel distance operates on unscaled branch lengths (original lengths) and another rescales by the mean.
We've been using the term "normalize" to refer to both the kernel normalization -- denominator = sqrt(k(t1, t1) * k(t2, t2)) -- and to rescaling branch lengths (
norm.mode
). From now on, "normalize" should be used exclusively in reference to the kernel distance, and the latter should be "rescaling" branch lengths.Either operation should not affect other distance measures. For example, it should be possible to have a distance function with two or more kernel distances, where one kernel distance operates on unscaled branch lengths (original lengths) and another rescales by the mean.