I think I may have figured out where we are diverging. When gathering data for your analysis, were you using the Global Infobase "estimates"? Or were you using the actual survey data from these countries? The two tend not to agree and the estimates seem to run high, for instance, in the 40s for the US, > 70 for micronesia, 35 for Mexico etc. That makes all the disagreements disappear from the obesity data, albeit not from the diabetes data. So I apologize if this generated confusion, it was not clear you were using the estimates rather than actual survey data from the methods.

However, I wonder, is it wise to use these estimates rather than actual survey data from the US such as that gathered by CDC? After all, when survey data is provided for each of these countries, it clearly contradicts the estimates, which I assume were calculated in 2005 if the citation is to be believed. Some of the countries appear to diverge by as much as 30%. I'm not able to find the methodology by which these estimates were calculated as the citation does not appear to exist in pubmed.

Two more things:

I found the 4th edition IDF atlas http://www.idf.org/node/2... and now figure 1 now makes a great deal more sense. It might have been clearer if in citation 6, the edition of the Atlas were noted. Also I feel the source of the obesity data is now clear, and from what I can tell, appears to be 7-year-old estimates of future obesity rates for 2010, not actual measured rates of obesity, which do suffer from infrequent updates. That these data don't correlate with actual measured survey data makes me question their use for such a representation, for many of the countries for such estimates, it's been decades since obesity rates were measured (eg Jamaica). They also are problematic as they are way outside of what seems plausible, or what have been measured by reliable public health agencies such as the CDC, but at least now I can reconstruct the data set It should be noted that in the newer diabetes data set, however, New Zealand is no longer showing a decrease, or divergence between obesity and diabetes, as the comparative prevalence is now back to 9%. And Sri Lanka is now much lower at 7%, is it possible the 2010 values were an aberration?

Second, the link you provided with the updated data has a permissions issue and remains inaccessible. I appreciate that this must have taken additional effort to provide more analysis, I'd like to be able to see it. I'm still hoping to create an identical reconstruction with the 4th edition atlas, as well as the newer diabetes data.