Distance Methods
Previous Top Next

Similarities and Dissimilarities (Distances)
Similarities measure the relatedness of a sample pairs, i.e. they measure how close to samples are to each other. Dissimilarities (distances) measure the number of differences between a pair of samples.

Absolute and relative distances
We can distinguish absolute distances from relative distances. For absolute differences the number of different positions is calculated. In the case of relative distances the absolute distance is normalized to the total number of positions that are compared.

Normalization
Distances are usually normalized (relative distances) to compensate different sample sizes and number of total positions. Thus, relative distances are numbers in the range of zero (no relatedness of two samples) to 1 (total match between a pair of samples).

Example
If 10 character positions are compared between two samples and both samples have two characters in common, they have an absolute similarity of 2 or a relative similarity of 0.2 (2 out of 10). The absolute dissimilarity or absolute distance is 8 (10 minus 2) or 0.8 (8 differing positions out of 10).

A large number of different distance measures are available and each distance measure has its own logic.

See
< Cluster analysis
> UPGMA method
> Neighbor-Joining method
> Principal coordinates analysis