Re: "date" or "boundary"?

[Christopher Ramsey]

Yes - you can - the meaning is different. If you use Boundaries, each phase is treated as a separate group - with a uniform prior for the length of each one. If you use dates for the intermediate events between phases then they will all be treated as a single group and with the events distributed randomly through that.

The main effect of this is if you have very different numbers of sampled dates from different phases. When using Boundaries, the model will still work on the assumption that all the phases have similar prior lengths regardless of the number of samples in each. If you use "Date" then the phases with more events will generally be longer than those with fewer.

Christopher

> On 23 Feb 2021, at 10:46, Hans-Christoph Strien <hcst...@gmail.com> wrote:
>
> Dear all,
>
> I learned that it is possible to use "date" instead of "boundary" for modelling boundaries between phases, but I couldn't find any information on the (dis)advantages of this choice - is there any paper I missed, or any other information?
>
> Thanks
> Hans-Christoph
>
> --
> 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/9a6b0d42-e211-47b4-9c66-255c4bb21b92n%40googlegroups.com.

[hcst...@gmail.com]

But then using "date" means the assumption, that the number of dates available correlates to the duration of phases - what in most cases is as uncertain as the assumption when using "boundary" that the duration of all phases is more or less equal. I made the experience, that with boundary the phases are usually modelled as of very different duration, mainly depending on the wiggles of the calibration curve (even in cases where we know that it is more or less equal), but in the only case I know where "date" was used, modelled durations correlate strongly to the number of dates per phase - which is most probably wrong.

Am I wrong assuming that the use of "date" is appropriate only in cases when we know that sampling was representative for the duration of the phases - what in reality usually is unknown? As e.g. with growing or shrinking population, even the number of features excavated per phase is no reliable proxy for time length, so we have no control for the basic assumption.

Hans-Christoph

[Christopher Ramsey]

Yes - they may end up being very different in length once the dates are included. Quite often we have by chance quite a few dates from some phases and very few from others - so it would not normally best to do it this way.

Where you might often want to use Date is if you have an excavation harris matrix and it provides a convenient way to specify the constraints between individual samples (short lived and not reworked) within the excavation sequence.

Best wishes

Christopher

> To view this discussion on the web visit https://groups.google.com/d/msgid/oxcal/bf95c0bf-978b-4f58-b228-c9e20e8f58can%40googlegroups.com.

[hcst...@gmail.com]

Is there any citation for these points? It would be useful for arguing why which decision was made.

And by the way: for every model it should be mentioned if "date" or "boundary" was used, and it should be argued why . Most archeologists don't even know the difference, and without explanation they won't recognize the difference - and even if yes, it would be necessary to discuss the reasons for the decision made by the authors - I have never seen something like this, but in my opinion it is an information absolutely necessary for readers.

Hans-Christoph

[Christopher Ramsey]

All the the key points of this are explained in:

http://dx.doi.org/10.1017/s0033822200033865

but obviously not every possible use of the different combinations.

Christopher

> To view this discussion on the web visit https://groups.google.com/d/msgid/oxcal/738671aa-ad15-4a06-9c77-a044de4a120bn%40googlegroups.com.