Selection Sunday

I need to send off around 50 samples to a colleague in England for oxygen isotope analysis. Why 50? Well, at 30£ a sample, that's what I can afford with what's left of the NSF and the Wenner-Gren. (If the pound falls further, though, perhaps I can add another sample or two.) So, as always, my conundrum is... which 50 teeth (people) do I send? Here are the stats.

My MNI (minimum number of individuals) is 212 from two sites, split about 70/30. Each site has a generally normal range of individual ages, but there is slight overrepresentation of males. For strontium, I tested every individual for whom I had a first molar (n=112). I have immigration data, then, on just over half of everyone I looked at. But I had to send off samples for carbon/nitrogen analysis, and I had budgeted money to test 50 individuals. I had to send them off before I got the strontium results back. So I decided on a more or less stratified sample proportional to the population of each of the two sites. I sent off samples from at least one individual in each 5- or 10-year age range, and for adults, I made sure to send at least one male and one female from each 10-year age range if present. What I ended up with, then, are data from 112 people for immigration and 52 of those people for diet, some of whom are probably immigrants.

Now, oxygen is going to help me figure out the strontium results. And since I need to send off first molar pieces, my sample is going to be 50 out of the 112 for whom I did strontium. I will likely have to omit a couple people from analysis because of lack of remaining enamel (which is destroyed in strontium analysis), but for the most part, I can choose any of the 112 for oxygen. My strontium values are more or less continuous: there are about half a dozen outliers in the high or low range. So, what's my sampling strategy? I obviously feel like I should test the individuals who are more or less clearly outliers. And then the rest can be pulled from the pool of middle-range values. I suppose a stratified sample of age and sex would be useful as well. My choice is: do I choose a new stratified sample based on the strontium values, or do I just send the same people for whom I got carbon/nitrogen results? If the former, I will have some individuals for whom I know Sr and O but not C/N and for whom I know Sr and C/N but not O. If the latter, I will have 50 people for whom I have Sr, O, and C/N, and 62 people for whom I only have Sr. I'm leaning towards this latter option. Or, I could just ignore these strategies and go back through the database looking for individuals I felt might be immigrants based on the phenotypic appearance of their bones while I was analyzing them. But basing ideas about immigrants solely on visual assessment of what looks "different" is probably not the best idea, especially since I only looked at a couple hundred people. I guess I'll play around tomorrow and see if it makes sense to send all the Sr/C/N people off for O.


Popular Posts