0
$\begingroup$

I have some data for the relatedness of different individuals, the distance between them and whether they belong to the same group.

I want to know: is there a significant effect of group identity on the effect of distance on relatedness. In other words, are group members more closely related than you would expect given their distance?

I found that it is possible to test whether there is a correlation between distance and relatedness using a mantel test, but haven't found anything similar that also allows inclusion of additional factors. Does anyone know if something like this exists?

$\endgroup$
1
$\begingroup$

I would run a simple linear regression of the following form:

$R_i = \beta _0 + \beta _1D_i + \beta _2G_i + \beta _3G_i\times D_i + \epsilon_i,$

where i would be a pair of two individuals $R_i$ - the measure of their relatedness $D_i$ - the distance between them, $G_i$ - a dummy variable which is one if both belong to the same group and zero if they dont. The interaction term $G_i\times D_i$ is what I would look at. If its coefficient $\beta _3$ is significantly different from zero and all the assumptions that underlie the OLS do not appear to be violated this would be evidence that group identity affects the relationship between relatedness and distance.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ The problem with this is that it doesn't account for the dependencies in the data (e.g. with 3 individuals i,j and k: Rij and Rik are not independent from Rjk), so it inflates the sample size. $\endgroup$ – unknown Dec 9 '19 at 21:41
  • 1
    $\begingroup$ @unknown if you are worried about depenedences in your variance-covariance matrix you can use HAC-standard errors, standard errors corrected for spatial correlation or bootstraped standard errors. You perhaps could take a look at this cameron.econ.ucdavis.edu/research/… $\endgroup$ – Grada Gukovic Dec 10 '19 at 12:12
  • $\begingroup$ Ah ok great, thanks! $\endgroup$ – unknown Dec 11 '19 at 15:19

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.