Lack of significance in AMOVA - (Jul/12/2011 )

Hi, I have had problems running an AMOVA on Arlequin. In the AMOVA I am observing allele frequency differences between four populations placed into two separate groupings (these correspond to two different species). The AMOVA indicates the largest degree of variation to exist between these two groups. However this is not statistically supported. From what I have read such a lack of significance within an AMOVA can be a consequence of having a small number of demes in your study. Can anyone therefore recommend a good alternative to an AMOVA in determining the significance of the variance components? I have heard suggestions that pairwise T-tests could be used.

Thanks

rob180 on Tue Jul 12 10:33:35 2011 said:

Hi, I have had problems running an AMOVA on Arlequin. In the AMOVA I am observing allele frequency differences between four populations placed into two separate groupings (these correspond to two different species). The AMOVA indicates the largest degree of variation to exist between these two groups. However this is not statistically supported. From what I have read such a lack of significance within an AMOVA can be a consequence of having a small number of demes in your study. Can anyone therefore recommend a good alternative to an AMOVA in determining the significance of the variance components? I have heard suggestions that pairwise T-tests could be used.

Thanks

hobglobin on Wed Jul 13 16:44:30 2011 said:

rob180 on Tue Jul 12 10:33:35 2011 said:

Hi, I have had problems running an AMOVA on Arlequin. In the AMOVA I am observing allele frequency differences between four populations placed into two separate groupings (these correspond to two different species). The AMOVA indicates the largest degree of variation to exist between these two groups. However this is not statistically supported. From what I have read such a lack of significance within an AMOVA can be a consequence of having a small number of demes in your study. Can anyone therefore recommend a good alternative to an AMOVA in determining the significance of the variance components? I have heard suggestions that pairwise T-tests could be used.

Thanks

I'm not sure if AMOVA is the appropriate analysis for phylogenetics, as the basic assumptions are not met when using different species (but not sure about that). I know it only for population genetics to statistically support a supposed grouping (which can be supported or not)....anyway I'd try a tree building method (NJ, max liklihood or parsimony) and then you have e.g. bootstrap values supporting the tree branching and separating the groups/populations, if the data are different enough.