Access

You are not currently logged in.

Access your personal account or get JSTOR access through your library or other institution:

login

Log in to your personal account or through your institution.

If You Use a Screen Reader

This content is available through Read Online (Free) program, which relies on page scans. Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.

Distance Methods for Inferring Phylogenies: A Justification

Joseph Felsenstein
Evolution
Vol. 38, No. 1 (Jan., 1984), pp. 16-24
DOI: 10.2307/2408542
Stable URL: http://www.jstor.org/stable/2408542
Page Count: 9
  • Read Online (Free)
  • Download ($4.00)
  • Subscribe ($19.50)
  • Cite this Item
Since scans are not currently available to screen readers, please contact JSTOR User Support for access. We'll provide a PDF copy for your screen reader.
Distance Methods for Inferring Phylogenies: A Justification
Preview not available

Abstract

The logic of inferring phylogenies from measures of the phenotypic distance between species is examined, and a statistical model constructed. Criticisms which have been levelled by Farris (1981) against methods of inferring phylogenies from distances are dependent on one particular interpretation of the entity being inferred. If we take the branch lengths on the tree to be expected distances rather than path lengths, the major criticisms of these methods lose their force. Under the expected distance view, there is no reason to abandon distance measures solely because they fail to satisfy the triangle inequality. The expected distance interpretation fits naturally into a statistical inference framework for inferring phylogenies. This in turn provides justification for some of the most widely-used criteria of goodness of fit between a tree and the observed distances. However, the statistical framework does emphasize the weakness of distance methods if the assumptions of additivity of distances among the tree and of independence of the measurement errors of the distances are not met. The possible failure of additivity and of independence seem to be the most serious problems with using distance methods. These assumptions are dubious for many kinds of data often analyzed by distance methods. The assumption of a molecular clock affects the details of the computations, but can be fit into the statistical framework without difficulty. Given the validity of the additivity and independence assumptions, a statistical test of the clock could be performed.

Page Thumbnails

  • Thumbnail: Page 
16
    16
  • Thumbnail: Page 
17
    17
  • Thumbnail: Page 
18
    18
  • Thumbnail: Page 
19
    19
  • Thumbnail: Page 
20
    20
  • Thumbnail: Page 
21
    21
  • Thumbnail: Page 
22
    22
  • Thumbnail: Page 
23
    23
  • Thumbnail: Page 
24
    24