Numeric characters: to range or not to range

[Mike Dallwitz]

                                                                  9 March 1995

> From: Glenn Hunt (Australian Museum)
> To: Mike Dallwitz
>
> One problem is making some characters numeric. Although it would inject more
> scientific rigor it might be too burdensome to the user. I see the potential
> for having two levels of characters: a character at one level might be
> useful for coarse discrimination, for example, the relative separation of
> the genital and anal plates might be coded as three character states which
> in fact allows separation of major classical groups of oribatids. The user
> could do this speedily at a glance. At another level a more precise numeric
> coding would allow fine discrimination, say at the species level, in certain
> taxa. I would envisage that all species would be coded for the character at
> the coarse level but only those where discrimination is required would be
> coded at the finer level.

Using numeric values directly, instead of in ranged form as multistate
characters, is not significantly more difficult for the INTKEY user, although
it is usually more work for the author to record actual numeric values.

As an example, consider a character to give the length of a wing. It might be
formulated as
    #1. length of wing/ mm/
or
    #2. length of wing/ 1. up to 10mm/ 2. 10-30mm/ 3. 30mm or more/

Consider a species where the wing length ranges from 17 to 19mm. This would be
coded as 1,17-19 or 2,2. The former requires the author to measure several
specimens, whereas the latter could be immediately entered just by glancing at
the specimens. The user identifying a specimen with a wing 17mm long could
also immediately enter 2,2, but could almost as easily estimate the length and
enter, say, 1,15-20. The former eliminates species which are always less than
10mm or more than 30mm, whereas the latter eliminates those which are always
less than 15mm or always more than 20mm - a substantially better result, with
little extra effort. The user has the option of getting a still better result
by making the measurement and entering 1,17, thus eliminating species which
are always less than 17mm or always more than 17mm.

For species in which the wing lengths are near the boundaries between the
ranges, for example, 9-11 or 11-13, both the author and the user will have to
make measurements to use the multistate form of the character, or accept
greatly reduced discrimination by entering both states: 2,1/2.

Most taxonomists tend to avoid using numeric characters, partly because it is
more work to record them, but also because they are not aware how good their
discriminating power can be, even when there is substantial intra-taxon
variability.

Recording actual numeric values also has advantages for classification, and
generation of printed keys. Most phenetic analysis programs can use the
numeric data directly. For programs which can't use the data directly, it can
be converted to multistate form (with the KEY STATES directive in CONFOR), and
the boundaries between the states can be chosen and experimented with AFTER
the data have been gathered.

Mike Dallwitz                                  Internet md at ento.csiro.au
CSIRO Division of Entomology                   Fax +61 6 246 4000
GPO Box 1700, Canberra ACT 2601, Australia     Phone +61 6 246 4075