Reported defects in Arlequin ver 2.000
(released 20 March 2000)
All versions (arlecore.exe)
The exact test of population differentiation does not take into account
the option "Infer haplotypes from distance matrix" in the General
Settings | Polymorphism control dialog box.
Therefore, it assumes that all the haplotypes you have entered are really
different to build the contingency table and does not recompute allele
frequencies. This has the effect that if you have entered raw haplotype data
with different identifiers for each haplotype in all populations, the
P-value of the exact test will be 1, as all permuted tables will have
exactly the same probability. Actually, this is not a real bug from our
side, but rather a feature of Arlequin. The exact test of population
differentiation lies on the comparison of allele or haplotype frequencies
among populations. It is the responsibility of the user to define correctly
what the alleles are, and to give them identical names in the different
populations where they appear. Thanks to Silvano Presciuttini for pointing
this out to us. (solved in ver 2.001)
The missing data character is considered as a real allele in the
estimation of haplotype frequencies from genotype data.
Therefore, frequencies of haplotypes with missing data at some loci
are estimated. This problem should be corrected in future releases. The
missing data allele will be correctly treated by assigning to it all
possible alleles when building the list of genotypes for each multi-locus
phenotype. This will cause the computations to be certainly longer. For the
moment, we would advise you to remove those individuals with missing data,
or remove the locus that presents too much missing data. Thanks to Alberto
piazza for pointing this out to us. (remains in ver 2.001)
The mean of the mismatch distribution does not correspond to the mean
number of pairwise differences computed under the molecular diversity
indices section.
This problem is due to the fact that for the mismatch distribution we always
compute the pairwise distance between sequences as the mere number of
pairwise differences, even though we allow the selection of an alternative
distance. Alternative distances should not be allowed because the mismatch
distribution is really the distribution of the number of observed
differences and not on the number of inferred differences (thanks to
Brigitte Pakendorf) (Solved in ver 2.001)
Unable to export Arlequin projects into other format.
This problem occurs when you have trailing blanks or tabs at ends of lines.
"Clean" the arlequin projects you want to convert beforehand. This
problem will be addressed in the next release (thanks to Douglas Prasher).
(Solved in ver 2.001)
Bad values for the confidence region of the observed mismatch
distribution.
The printed values of the lower and upper bounds of the confidence region
for the observed mismatch are wrong, and quite random. This is due to the
fact that we did not sort the values obtaind by the bootstraps before
printing. This will be corrected in the next release of Arlequin (thanks to
Victor DeFilippis). (Solved in ver 2.001)
Mac version
There is currently no way to end the console core program
launched by the Java interface. The user has to close the
console window by typing apple-Q.
The java interface also
cannot launch automatically the browser interface after
the end of the computations. This has to be done manually
by the use from the interface.
Manual
Erratum on p. 85. The correct equation for the probability of the observed
table should be
(Thanks to Alicia and Zé)
Erratum on p. 99. Correct expectation for Nei's genetic distance DA
is
(Thanks to Alicia)
Erratum on p. 97. proper references for the computation of the
Minimum Spanning Tree can be found under section 7.1.2.9 on p. 83. and not
under section 7.1.8.
(Thanks to Brigitte Pakendorf)
The batch file entitled run-jre.bat is actually named arlequin.bat
in your Arlequin directory.
Bugs of Arlequin version 2.000 actually fixed in
version 2.001
The automatic creation of Arlequin projects wrote key words in a wrong
order, causing some projects to be impossible to read by the Arlequin
interface.
Genepop conversion utility has been made more compatible (but there are
still some problems)
Minimum spanning tree structure generated in NEXUS format could not be
read if sample names contained white spaces.
Genotype likelihoods were not correctly computed in the genotype
assignment procedure.
External distance matrices could not be read if the key word LabelPosition=LINE
was not explicitly stated.
Fct p-value was not correctly estimated for the locus-by-locus Amova.
Computed heterozygosity was not correctly estimated when haplotypes with
zero frequency were present.
AMOVA: P-values were generally not well estimated for small sample sizes
or small number of populations. This was because of rounding errors when
comparing observed and randomized F-statistics.
Mismatch distribution: Upper and lower bounds for the mismatch
distribution were not correctly computed
Comments within groups structure did not allow samples to be read and
project to load correctly.
Computations of genetic distances taking into account population size
difference caused Arlequin to crash when one population was monomorphic.
The mean of the mismatch distribution did not always correspond to the
mean number of pairwise differences computed under the molecular diversity
indices section. This was due to the fact that one had the impression that
one could select different distance for the mismatch, while only the number
of pairwise differences is always computed for the mismatch.
Unable to export Arlequin projects into other format. This problem occurred
when when there were trailing blanks or tabs at ends of lines.
