--- In ANEemail@example.com
, BjÃ¶rn Lindborg <bjorn07se@...> wrote:
> It's interesting to note that the high and low schools in Palestine
> Iron Age chronology (Mazar and Bronk Ramsey vs. Sharon and Gilboa)
> arrive at their resp. positions by analysis of basically the same
> raw 14C dates.
> [I think the Radiocarbon article by Mazar and Bronk Ramsey (Vol.
> 50:2 pp. 159-180; August 2008) is not yet available on the web.]
I received Radiocarbon volume 50:2 today, and went over the Mazar and Ramsey publication. That after going through most of the analysis published so far. It's absurd at best, both Finkelstein and Mazar use largely the same data and appear to reach opposite conclusions. It's a problem, but I think it can be solved.
Bayesian analysis is always problematic, because you need to be very careful about the assumptions you make in your models. They both use more or less the same models but varying datasets, but still - different models lead to different results.
There's also an information problem. Radiocarbon analysis has a lot of noise that makes pinpointing dates accurately a problem. While each new sample adds a little bit of information in that aspect, but clearly a lot more data is required to be able to zero in on the correct decade of the Iron I to Iron IIa shift.
This information problem leads to relatively unstable mathematical solutions. That's why changing just a few data points in the data sets causes apparant dramatic changes in the results.
But the real problem is none of the above. The real problem is that the models are all more or less right (because they're all very similar). The problem is interpreting the results:
If you look at the Mazar and Ramsey article mentioned above, in page 173 they have several graphs depicting the results of the four different Bayesian models they used on the data. The first row - the A models, is their attempt to reproduce Finkelstein's results. They succeeded. Look at the graph for model A2. It is quite obvious the 'one sigma interval' is somewhere between 950 and 920 BC, supporting the low chronology. However, in the same graph, the probability of the Iron I/IIa shift occuring somewhere BEFORE 975 BC is somewhere between 10% and 15% (it's hard to say exactly how much without running the model and looking at the number). So even Finkelstein's own analysis attributes a real non-neglible probability for the high chronology. While 70% is more than 10%, it doesn't prove the 10% are wrong.
It works the other way around, too. The bottom row contains graphs for Model D - one of the models that supports the high chronology according to the authors. There, the probability of the Iron I/IIa shift occuring after 950 BC seems to be around 25%.
So, if I got the jist, after a decade of crunching numbers and trying similar but different statistical models, everybody reaches the same conclusion - both chronologies are likely. I can't see which one is more likely based on the numbers I've seen, but even if one is a bit more likely than the other, it doesn't prove anything. At this point, statistics just can't help solve this debate.