I have binomial frequency data for an allele associated with populations living in mountainous. These mountains run north to south where sites are nearly fixed for this allele and lowland sites to the east that lack the allele. At the south end of the mountains is a hybrid zone. Despite the mountains continuing south through the hybrid zone, the allele frequency shifts from nearly fixed to nearly absent. This suggests that despite the assumed benefit for this trait in the mountains, that there is a barrier for it in the hybrid zone preventing its spread south in the mountains.

I have used a simple glm approach in R to identify that altitude is a significant predictor of the allele frequency north of the hybrid zone. If I include hybrid zone sites in the model, altitude is less significant. If I add a hybrid zone dummy variable to the model and it is also significant. I would like to find a statistical approach to identify these "outlier" mountain sites in the hybrid zone both statistically and graphically.

  • 1
    $\begingroup$ One idea to explore might be the inclusion of Latitude in your model. Another might be to look at Longitude, though if the mountains are in west then this could be highly correlated with altitude. $\endgroup$ – Henry Mar 31 '12 at 12:12
  • $\begingroup$ Yes, putting latitude and longitude in the model. They are both informative, but largely confirm what we can see by visually looking at the data (just like altitude). This still leaves me with the question, how can I identify sites in the mountains that do not have the "expected" allele frequency of mountain sites? $\endgroup$ – Keith Larson Mar 31 '12 at 12:27
  • $\begingroup$ If I analyze all the data from the hybrid zone and the mountains to the north, including latitude in the model, I would expect a significant positive effect with latitude if there is a north-south gradient. This would suggest that the hybrid zone in the south (or some other factor yet considered!) is a barrier to the spread of this allele southwards in the mountains. I would still like to identify sites with lower than "expected" frequencies. Thanks Henry for making me think about my data differently! $\endgroup$ – Keith Larson Mar 31 '12 at 13:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.