Miscel.lania Zool oqi ca 22.

1 (1999)
Tables of significant values of
Jaccard's index of similarity
R. Real
Real, R., 1999. Tables of significant values of Jaccard's index of similarity. Misc. Zool., 22.1: 29-40
Tables ofsignificant values oflaccard's index ofsimilarity- Two statistical tables of probability values for
Jaccard's index of similarity are provided. Table 1 is t o substitute a previously published table and is
applicable when any possible distribution for the N elements in both OTUs is considered. Tables 2 and 3
are applicable when fixing a set number of total attributes in each OTU.
Key words: Statistical tables, Jaccard's coefficient, Association analysis
(Rebut: 26 199; Acceptaci condicional: 2 VI 99; Acc. definitiva: 29 V199)
Raimundo Real, Depto. de Biologa Animal, Fac. de Ciencias, Uniil: de Mlaga, 29071 Mlaga, Espaa
(Spain).
e-mail: rrgimenez@uma.es
ISSN: 021 1-6529 O 1999 Museu de Zoologia
Real
Jaccard's similarity index (JACCARD, 1908) con-
siders t he similarity between t wo OTUs (Op-
erational Taxonomic Units) as t he number
of attributes shared divided by t he t ot al
number of attributes present i n either of
them. Jaccard's index may be expressed as
follows:
where A i s t he number of attributes present
i n OTU a, B is t he number of attributes
present i n OTU b, and C i s t he number of
attributes present i n bot h OTUs a and b. The
number of attributes present in either of t he
OTUs (N) i s given by A+B-C.
Jaccard's index i s widely used in regional-
ization and species association analyses, al-
though its probabilistic basis i s not usually
taken i nto account (REAL et al., 1992). How-
ever, it i s feasible t o determine al1 the possible
distributions of N attributes in any of t he
categories A, B or C of the previous formulae
for each value of N, and so an exact random-
ization test (SOKAL & ROHLF, 1981, p. 788) can
be performed t o determine whether an ob-
served value of J is significantly different from
those expected at random.
BARONI-URBANI (1980) studied Jaccard's simi-
larity index from a statistical point of view and
obtained a statistical table of associated prob-
abilities; this table i s applicable when any pos-
sible distribution for the N elements in both
OTUs is considered, and thus free reversibility
of the attributes is allowed. REAL & VARGAS
(1 996) modified these probabilities in t wo ways:
a) they amended a fl aw in the formulae used
by BARONI-URBANI (1980) t o obtain the prob-
abilities associated t o the similarity index; b)
they obtained another set of formulae, first
mentioned in REAL et al. (1992), that were
applicable when fixing a set number of total
attributes in each OTU, where the attributes
are considered as irreversible. The latter set of
formulae are t o be prefered when the number
of attributes present in the t wo OTUs com-
pared are considered as necessarily different,
as the number of species present in t wo is-
lands of very different surfaces.
However, t he calculus of t he probabilities
associated t o Jaccard's index using t he for-
mulae i n REAL & VARGAS (1996) involves t he
determination of al1 t he possible outcomes
of t he distribution of N attributes i n t he t wo
OTUs, and this takes an enormous amount
of ti me even for a modern computer, spe-
cially when fixing t he number of attributes
i n each OTU, so rendering these formulae as
of l i ttl e practica1 value. It i s therefore neces-
sary t o provide some tables so that these
probabilities can be easily applied according
t o t he assumptions of t he researcher.
Table 1 shows the lower and upper critical
values of Jaccard's index wi t h the probability
levels 0.05, 0.01 and 0.001, when any possi-
ble distribution for t he N elements in the
t wo OTUs i s considered. In this case the prob-
abilities associated wi t h Jaccard's index de-
pend only on the total number of attributes
present in either of the t wo OTUs being com-
pared (N). Table 1 must then substitute the
statistical table in BARONI-URBANI (1980).
Tables 2 and 3 show the lower and upper
critical values of Jaccard's index, respectively,
wi th the probability levels 0.05,0.01 and 0.001,
when fixing a set number of total attributes
i n each OTU. In this case t he probabilities
associated wi t h Jaccard's index depend on
the total number of attributes present i n ei-
ther of the t wo OTUs compared (N) and on
t he number of attributes in t he OTU that
displays the lowest number of attributes (B).
Tables 2 and 3 are considerably shorter
than t he correspondent tables where t he
number of attributes of each OTU (A and B)
i s considered instead. However, given that N
is different f or each pair of OTUs compared
and t hat many statistical programs do not
provide the values of N associated t o each
value of Jaccard's index, these tables may
require t he time consuming activity of count-
ing t he number of attributes shared by t he
t wo OTUs (C) i n order t o infer N. This may be
avoided calculating N from the Jaccard's value
(J) and t he values of A and B i n t he fol l ow-
ing way:
Resumen
Tablas de valores significativos para el ndice
de similitud de Jaccard
En el presente trabajo se aportan dos tablas
estadsticas de probabilidades asociadas al
ndice de similitud de Jaccard. La tabla 1
Miscel.lania Zoologica 22.1 (1999)
3 1
sustituye a una tabla publicada previamente
y es aplicable cuando se permite cualquier
distribucin de los N elementos en los dos
OTUs. Las tablas 2 y 3 son aplicables cuando
se fija el nmero de atributos presentes en
cada OTU.
References
BARONI-URBANI, C., 1980. A statistical table f or
t he degree of coexistence between t wo
species. Oecologia, 44: 287-289.
JACCARD, P., 1908. Nouvelles recherches sur la
distribution florale. Bull. Soc. Vaud. Sci.
Nat., 44: 223-270.
REAL, R. & VARGAS, J. M., 1996. The probabilistic
basis of Jaccard's index of similarity. Syst.
Biol., 45: 380-385.
REAL, R., VARGAS, J. M. & GUERRERO, J. C., 1992.
Anlisis biogeogrfico de clasificacin de reas
y de especies. In: Objetivos y mtodos
biogeogrfi COS. Aplicaciones en Herpetologa.
Monogr. Herpetol., 2: 73-84 (J. M. Vargas, R.
Real & A. Antnez, Eds.). Asociacin
Herpetolgica Espaola, Valencia.
SOKAL, R. R. & ROHLF, F. J., 1981. Biometry. 2nd
ed. Freeman, New York.
1
32 Real
Miscel.lania Zoologica 22.1 (1999) 33
34 Real
Miscel.lania Zoologica 22.1 (1999) 3 5
-
36 Real
Miscel.lania Zoologica 22.1 (1999) 37
Real
Miscel.lania Zoologica 22.1 (1999) 39
40 Real
1