Biological distance was a widely used method to reconstruct relatedness among populations. However, as you noted, other developmental factors are involved on the establishment of the bone morphology. Whether you can use this method probably depends on your objective, the geographic range of your samples and the age of your populations. I believe that it would be ok (though imprecise) to infer relatedness of population with more than 10.000 years along the globe by measuring their skull. But measuring long bones of, for example, modern indigenous to south america, I would hardly believe that it could bring consistent results. In this case I would recommend the use of population genetic tools, that will surely be much more precise on inferring relationship among populations.

Before the advent of easy genotyping, phenotypic distances (such as you suggest) have been used to estimate relatedness between populations. However, a 1997 paper by Burstin and Charcosset (see link) showed that there exists a triangular relationship between genetic and phenotypic distances: Short genetic distances are associated with short phenotypic distances, but long genetic distances are associated with a wide range of phenotypic distance. In other words: Phenotypic distances are not all that predictive of genetic distances.

I agree with @Pedro: If possible use some type of genetic markers to infer relatedness or genetic distance.

http://www.nature.com/hdy/journal/v79/n5/pdf/hdy1997187a.pdf

Joe Felsenstein developed methods for inferring phylogeny from quantitative characters. You'll find these in his PHYLIP package.

John Relethford studied the problem from an intraspecific perspective, looking at relationships among human populations. I'm not sure which of his papers is the best reference: the ones below are my best guesses. You're looking for his method of "R matrices".

@article{relethford1994craniometric,

title={Craniometric variation, genetic theory, and modern human origins},

author={Relethford, Dr and John, H and Harpending, Henry C},

journal={American Journal of Physical Anthropology},

volume={95},

number={3},

pages={249--270},

year={1994},

publisher={Wiley Online Library}

}

@article{relethford1994craniometric,

title={Craniometric variation among modern human populations},

author={Relethford, John H},

journal={American Journal of Physical Anthropology},

volume={95},

number={1},

pages={53--62},

year={1994},

publisher={Wiley Online Library}

}

Given the vast number of SNPs associated with stature, and presumably, long bone length in humans, one would expect genetic distances to be very highly correlated with phenoytpic distances for long bone length.  I don't know of recent studies that have addressed this, though. 

Benedikt, why would that be expected? If  the SNP effects on  bone length are additive (more or less), the distribution of any two samples of a genotype will show (on average) quite some difference, while the resulting bone lengths is distributed normally over a limited range. Chances are that the two genotypes show very similar bone lengths, while the difference in genotype is considerable.

Moreover, you could think of two genotypes: A is homozygous for allele 1 for every odd SNP and homozygous for allele 2 in every even SNP, and for B it is the other way around. Genotype A and B have zero similarity and hence maximimum distance. But provided the SNP-effects are distributed more or less evenly, the resulting phenotype will be the same.

That is what the paper by Burstin and Charcosset showed: If two (aggregate or mean) genotypes are alike, the phenotype will be similar as well. If two genotypes are very different, the phenotype could be very different, but it could also be very similar. The correlation between similarity/distance between genotypes and phenotypes vanishes with increasing distance.

Herwin, That is an excellent point, especially for a low dimensional trait like long bone length where multiple genotypic states would be expected to have the same mean phenotype.  My initial reasoning was based on the fact that such a large proportion of the genome influences stature.  For a high dimensional trait like craniofacial shape, I think that reasoning might make more sense. However, I agree with your argument above.

Benedikt, perhaps the correlation between genotypic and phenotypic distance is more constant for higher dimensional traits. I've no real experience with those, except when they're broken down into one-dimensional constituent parts.

However, I would think that if the high dimensional trait can be expressed as a function of several one-dimensional traits, the same reasoning by B&C holds: On average the phenotypic distance between constituent traits will be a fraction of the genetic distance. Hence the phenotypic distance of a higher dimensional trait will also under-estimate the 'real' distance considerably, by virtue of it being a function of the one-dimensional traits we know are underestimating the distance.

There are a variety of methods that could be used for such a study, many of them can be found in Hammer and Harper 2006 "Paleontological data analysis". I would suggest canonical variates for multiple 'groups', or discriminant function analysis for two 'groups'. Be aware that the number of variables you can study is equal to the smallest population sampled in one of your 'groups'.

Also, you have to be careful that even if you do find a statistically significant separation this is not due to some other biological pressure, such as nutrition, or average age sampled etc.

Ryan Harrod has used post-cranial measurements to do some biodistance comparisons. His study might provide some background literature although I'd avoid post-cranial elements and also try to use periosteal reaction patterning as a control to suggest that health disparities within comparative groups was small (admittedly, this is still a lacking control given the "osteological paradox").

The citation of the study I reference is: 

Ryan P. Harrod. 2011. Phylogeny of the Southern Plateau: An Osteometric Evaluation of Inter-Tribal Relations. Journal of Comparative Human Biology HOMO 62(3): 184-201.