Arlequin Significant Fst Values For Population Sizes N=1, Do Not Match Genodive Fst Values

Question

[luckylion07]

Hello,

I did calculate Fst values with 3 microsatellite makers for 29 populations using Arlequin 3.5.1.3 (Population pairwise Fst values: Compute F-statistics on haplotype frequencies only). However I have a population "A" with only one individual. Fst values of "A" compared to other populations shows significant p-values (p<0.05). How is that possible? I did check the input file, no problem there. From what I learned, sample sizes of n=1 can never be significant.

Also I calculated Fst values with GenoDive, wich is leading to slightly different Fst values (mean diference 0.0151, sd=0.0699).

Can you please confirm or comment on, if significant values for population sizes n=1 are valid?

If some one has a suggestion for a alternative program that can calculate different varieties of Fst with p-values i would be very happy! Thank you for your time!

here a part of the input file for arlequin:

[Profile]
  Title = "..." 
  NbSamples = 29
  GenotypicData = 1
  LocusSeparator = WHITESPACE
  GameticPhase = 1  # = unknown gametic phase
  MissingData = '?'
  CompDistMatrix = 1
  DataType = STANDARD

[Data]

[[Samples]]
    SampleName = "BB" 
    SampleSize = 1
    SampleData = {
    BB_01   1  175  300  168
               175  300  168
    }

    SampleName = "E03" 
    SampleSize = 20
    SampleData = {
    E03_01  1  177  300  164
               177  300  168
    E03_02  1    ?    ?  164
                 ?    ?  164
... continues

Answer 1

[Zev.Kronenberg]

Fst is a score. You permute alleles to get an empirical P-value. You can run Fst for an individual, but you shouldn't invest too much faith in the output. With only one individual it becomes almost binary in nature.

The P-value is correct, but what does that mean? Not much.