18-04-2024, 12:39 AM
Hello Andrew,
the calibration curve comes with its own uncertainty, and that is the main contributor to the overall uncertainty of the result. In fact, already combining just any three of the samples gives essentially the same result as using all four.
Effectively, combining the samples just reduces the uncertainty of the measurements (fractional C14 content), but the uncertainty from the calibration curve is not affected and stays with us. It cannot be, and is not reduced by using more samples.
This curve (and its standard deviation) has been established over decades by the global communicty of radio-carbon centres.
Greg Hodgins used the 2004 issue of the curve. I have also tried using the 2009 and 2013 issues, and the difference was exactly zero. (I found these on the net. I think one can still find them, e.g. here:
You are not allowed to view links. Register or Login to view. ).
One other possibility why the combination could be invalid is if more than one of the folios are, by chance, from the same skin. At least partially invalid.
However, as I wrote above, using three pages out of four already gives the same result, so that shows that this is in fact not an issue.
Now of course, (and I mentioned this already to Greg Hodgins - and he agreed), there could be other types of systematic errors that are not affected by using several samples. He was reasonably confident in the result, however. Of course he should be. He is working in his own discipline, and has a lot more background knowledge. Anything that I (or anyone else here) could come up with, would already have been well established.
Anyway, to check this out, I ran my own version of the calibration with a scale factor (1.5, 2, 2.5) on the sigma of all the samples.
(I am sure that I wrote this somewhere, but no idea when or where).
The outcome was NOT that the same scale factor applied to the result. Far from it. This is again because the uncertainty is dominated by the uncertainty of the calibration curve.
Effectively, the upper limit hardly changes, and the lower limit would go down by a couple of years, also clearly below 1400. Mainly, the second maximum around 1340 would come back into the 95% range.
the calibration curve comes with its own uncertainty, and that is the main contributor to the overall uncertainty of the result. In fact, already combining just any three of the samples gives essentially the same result as using all four.
Effectively, combining the samples just reduces the uncertainty of the measurements (fractional C14 content), but the uncertainty from the calibration curve is not affected and stays with us. It cannot be, and is not reduced by using more samples.
This curve (and its standard deviation) has been established over decades by the global communicty of radio-carbon centres.
Greg Hodgins used the 2004 issue of the curve. I have also tried using the 2009 and 2013 issues, and the difference was exactly zero. (I found these on the net. I think one can still find them, e.g. here:
You are not allowed to view links. Register or Login to view. ).
One other possibility why the combination could be invalid is if more than one of the folios are, by chance, from the same skin. At least partially invalid.
However, as I wrote above, using three pages out of four already gives the same result, so that shows that this is in fact not an issue.
Now of course, (and I mentioned this already to Greg Hodgins - and he agreed), there could be other types of systematic errors that are not affected by using several samples. He was reasonably confident in the result, however. Of course he should be. He is working in his own discipline, and has a lot more background knowledge. Anything that I (or anyone else here) could come up with, would already have been well established.
Anyway, to check this out, I ran my own version of the calibration with a scale factor (1.5, 2, 2.5) on the sigma of all the samples.
(I am sure that I wrote this somewhere, but no idea when or where).
The outcome was NOT that the same scale factor applied to the result. Far from it. This is again because the uncertainty is dominated by the uncertainty of the calibration curve.
Effectively, the upper limit hardly changes, and the lower limit would go down by a couple of years, also clearly below 1400. Mainly, the second maximum around 1340 would come back into the 95% range.