(adapted from the original Genepop 4.0 documentation written by Francois Rousset)

For this option the null hypothesis
Ho is: "*Genotypes at one locus are independent from genotypes at the other locus*". For a pair of diploid loci, no assumption is made
about the gametic phase in double heterozygotes. In particular, it is not inferred
assuming one-locus HW equilibrium, as such equilibrium is not assumed anywhere
in the formulation of the test. The test is thus one of association between diploid
genotypes at both loci, sometimes described as a test of the composite linkage
disequilibrium (Weir, 1996, p. 126{128). For a haploid locus and a diploid one, a
test of association between the haploid and diploid genotypes is computed (there is
no concern about gametic phase in this case). This makes it easy to test for cytonuclear
disequilibria. For a pair of loci with haploid information, a straightforward
test of association of alleles at the two loci is computed.

The default test statistic is now the log likelihood ratio statistic (G-test). However one can still perform probability tests (as implemented in earlier versions of Genepop) by using the Gametic Disequilibrium Test = Probability setting on the input form.

For a given pair of loci within one sample, the relevant information is represented
by a contingency table looking e.g. like

GOT2 1.1 1.3 3.3 1.7 3.7 EST ______________________________ 1.1 1 1 0 0 1 3 1.2 16 6 1 3 2 28 ______________________________ 17 7 1 3 3 31

for two diploid loci (1.1, etc., are the diploid genotypes at each locus). Contingency tables are created for all pairs of loci in each sample, then a G test or a probability test for each table is computed for each table using the Markov chain algorithm of Raymond & Rousset (1995a). The number of switches of the algorithm is given for each table analyzed.

**OUTPUT**. Results are returned via the
web browser (and/or you can also have the results emailed to you - this may be
advisable for users with large data files as the computations may take some time).
Three intractable situations are indicated: empty tables ('No data'), table with one row or one column only ('No
contingency table'), and tables for which all rows or all columns marginal sums
are 1 ('No information'). For each locus pair within each sample, the unbiased
estimate of the P-value is indicated, as well as the standard error. Next, a global
test (Fisher's method) for each pair of loci is performed across samples.

All contingency tables (for *n* loci and *p* populations, there are *p*.*n*.(*n*-1)/2 tables)
are available through sub-option 2. Tables will be returned to you via the web
browser (and/or email).

**Please report any problems or bugs that you encounter with the web version
of this option to
Eleanor Morgan**

[Genepop Option 2] [Genepop Home Page] [Bibilography]