T value

[Andreu Monforte-Barberan]

Dear colleagues,

I am writing to you because I have a question about the value of T obtained by applying the Combine function. How is this value calculated when combining non gaussian distributions? How is the significance threshold of 3.8 calculated?

Thank you very much.

Andreu Monforte-Barberán

[Christopher Ramsey]

Dear Andreu

That is a good question. For R_Combine and C_Combine this is just the Ward and Wilson methodology:

https://doi.org/10.1111/j.1475-4754.1978.tb00208.x

with the threshold value calculated from the degrees of freedom.

The quoting of a value in Combine is because the normal use of Combine is actually a special case of a D_Sequence where there is no gap between the measurements. It turns out there is a direct relationship between the maximum probability in the overlap integral and the chi-squared value as outlined in Bronk Ramsey et al 2001:

http://dx.doi.org/10.1017/s0033822200038248

So for a D_Sequence the minimum chi-squared - is the least squares fit to the curve. This is quoted for the D_Sequence and also for the Combine for this reason. For the combine - it is the equivalent of combining the dates after including the error in the calibration curve and the maximum probability fit point - so the T value will be lower than if you use R_Combine. This is consistent with the reason why you should use R_Combine rather than Combine for dates which are all from the same event.

If you use Combine for normal distributions - for example in place of C_Combine - the T value will be the same because there is no calibration uncertainty - this is then just as in Ward and Wilson.

Best wishes

Christopher

> --
> You received this message because you are subscribed to the Google Groups "OxCal" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to oxcal+un...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/oxcal/51e6b8f1-e3e4-452f-a014-f96cc0b42dadn%40googlegroups.com.