How to use Fst_like and collate.

    PRODUCING A DATA FILE

Your data should be in a text file organised as in the example dataset
"data.", which is reproduced below:
---------------------------------------------------------------------
3 populations
3 loci

10 vwa alleles (12 through 21)
   0    0    14     14    11      33     22     10    2    0  %pop1
   0    1    11     11    35      42     36     10    0    0  %pop2
   1    3    50     89    137     194    137    60   10    1  %pop3
7 tho1 alleles (5 through 11)
 0   29     10    10     28    27    0  % pop1
 0   41     15    23     37    32    0  % pop2
 0   202    103   85     127   165   0  % pop3

3 test alleles
0 0 0
0 0 0
0 0 0
---------------------------------------------------------------------

Notice that the first two lines lines of the file (apart from blanks)
are the number of populations and loci.

The Data for each locus then follows as a series of matrixes.

Each matrix is preceeded by a line giving the number of
alleles.  There is a row for each population and a column for each allele.
Missing data for any one population is given as a series of zeros.

The text (eg "% pop3") is optional and can be included at the end of
any line which contains numbers.

         ESTIMATING Fst

You require the programs "Fst_like" and "collate" which you can obtain
as pascal source files (.pas) or as .exe files which will run on a PC.

You should run Fst_like first.  It will generate a .out file for each combination
of locus and population.  These are text files containing two columns:
Fst value and Log likelihood.
They can be used to plot log-likelihood curves.
They may be examined or read into a graphics package of your choice and plotted.
This may enable you to spot anomalous populations or loci.

You can then combine information across loci and/or across populations using
the program collate.  This should be run from the directory containing
the output files from Fst_like.  It gives you the option to leave some
loci or populations out of the analysis, or even to run it for only one
population-locus combination if you want an individual probability curve
(rather than a likelihood curve).

The output is in the form of 3 columns: Fst value, Probability density, cumulative probability.
The first two can be plotted to give a probability density curve for Fst.
The largest density is opposite the Maximum Likelihood value.  The
cumulative density can be used to calculate credible intervals for Fst
eg the 0.025 and 0.975 points give the 95% interval. (assuming a uniform prior).

Feedback to R.A.Nichols@qmw.ac.uk please.