P_Sequence age-depth model output

[Robert Kelly]

I'm a relative newcomer to age-depth modelling. I can see under the Raw Data table, in the Depth Model box, where I can set the depth (or elevation) and increment limits of the model. I then click on the <<Depths button and the modelled ages by depth increment appear in the opposite box. Those data include the 95% confidence limits of the age model for each increment of depth, defining an age range for each increment of elevation. My question is: Are those age ranges bracketed by the confidence limits defining a normal distribution, a uniform distribution, or, more likely, a 'funky' distribution similar to what one sees in a calibrated radiocarbon date. And, if the latter, how inappropriate might it be, in a  fairly tight age-depth model (narrow packets of probability between radiocarbon dates)  to treat the distribution at each elevation increment as a normal or uniform distribution? 

The program produces modelled distributions with probability assigned to the different portions of age ranges for each radiocarbon dates employed in the model, but not, to my knowledge, for the modelled distributions at elevations between those dates, or am I wrong?  Thank you. Bob Kelly

[Christopher Ramsey]

Dear Bob

The values reported in the right hand table are based on probability distributions inferred at the points included in the model and at interpolated points. The ranges there will be highest probability density ranges. The interpolation distributions can be multi-model here (not normal or uniform distributions).

The interpolation algorithm does not try to do anything complicated with this - it just linearly interpolates the ranges from the right hand table into the left hand table at the requested values for convenience. Because the P_Sequence does, in its default mode, already interpolate between dated points there are generally enough points to give a reasonable interpolation. To look at it another way - this process is effectively equivalent to reading the values off the age-depth model plot - where straight lines are drawn between the range end points.

Best wishes

Christopher

> On 8 Dec 2025, at 23:14, 'Robert Kelly' via OxCal <ox...@googlegroups.com> wrote:
>
> I'm a relative newcomer to age-depth modelling. I can see under the Raw Data table, in the Depth Model box, where I can set the depth (or elevation) and increment limits of the model. I then click on the <<Depths button and the modelled ages by depth increment appear in the opposite box. Those data include the 95% confidence limits of the age model for each increment of depth, defining an age range for each increment of elevation. My question is: Are those age ranges bracketed by the confidence limits defining a normal distribution, a uniform distribution, or, more likely, a 'funky' distribution similar to what one sees in a calibrated radiocarbon date. And, if the latter, how inappropriate might it be, in a fairly tight age-depth model (narrow packets of probability between radiocarbon dates) to treat the distribution at each elevation increment as a normal or uniform distribution?
>
> The program produces modelled distributions with probability assigned to the different portions of age ranges for each radiocarbon dates employed in the model, but not, to my knowledge, for the modelled distributions at elevations between those dates, or am I wrong? Thank you. Bob Kelly
>

> --
> 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 visit https://groups.google.com/d/msgid/oxcal/3ea813ec-a5bd-45fc-b0cd-8750bca48775n%40googlegroups.com.

[Robert L. Kelly]

Dear Christopher,

Thank you very much. This is what I expected to be the case. And so my new question is: can one access the interpolated probability distribution at a given elevation? My goal is to use those distributions to select random dates for artifacts at the given elevation. 

Thank you,

Bob

Get Outlook for iOS

---

From: 'Christopher Ramsey' via OxCal <ox...@googlegroups.com>
Sent: Tuesday, December 16, 2025 10:37:32 AM
To: OxCal group <ox...@googlegroups.com>
Subject: Re: P_Sequence age-depth model output

 

◆ This message was sent from a non-UWYO address. Please exercise caution when clicking links or opening attachments from external sources.

To view this discussion visit https://groups.google.com/d/msgid/oxcal/5FB4CC31-4EFE-4B46-8D5D-CA07A525D557%40arch.ox.ac.uk.

[Christopher Ramsey]

Dear Bob

Assuming you don't have too many such points you can just include and undated event like:

Date("Name"){z=1.23;};

to give that distribution within the model. The program does not generate distributions for every possible depth and it is not really possible to derive and interpolated distribution directly without redoing the modelling process.

Best wishes

Christopher

> On 17 Dec 2025, at 16:24, 'Robert L. Kelly' via OxCal <ox...@googlegroups.com> wrote:
>
> Dear Christopher,
> Thank you very much. This is what I expected to be the case. And so my new question is: can one access the interpolated probability distribution at a given elevation? My goal is to use those distributions to select random dates for artifacts at the given elevation.
> Thank you,
> Bob
>
> Get Outlook for iOS

> To view this discussion visit https://groups.google.com/d/msgid/oxcal/MWHPR05MB34865520DDF40CDA212AA0DFDDABA%40MWHPR05MB3486.namprd05.prod.outlook.com.

[Robert L. Kelly]

Christopher?

I’m guessing that 2000 would be “too many”?

Bob

Get Outlook for iOS

---

From: 'Christopher Ramsey' via OxCal <ox...@googlegroups.com>

Sent: Wednesday, December 17, 2025 4:33:49 PM

To view this discussion visit https://groups.google.com/d/msgid/oxcal/F85A6260-C797-42DC-963A-E49CC7822A32%40arch.ox.ac.uk.

[Christopher Ramsey]

Dear Bob

Yes - I think that would be hard to manage.

The interpolation process tries to even up the density of points at which modelling is performed - so there should always be a distribution near by any point - just not actually at that point.

The other thing you could use if the ensembles of output models - this generates ensembles of possible age-depth models. These can be interpolated to all the depths you want. This will give you correlated samples of the possible model outputs for each level.

For this to work you need to set up the ensembles for the run using [Tools > Options > Advanced] where you can set say 100 outputs and then if you go to the age-depth plot and select [Edit > Format > Options] set the ensemble number to be shown on the plot. Once set the raw view will also generate a column for each of these ensemble curves.

Note that for ensembles to work you have to use the interpolation option in P_Sequence.

Best wishes

Christopher

> On 18 Dec 2025, at 19:23, 'Robert L. Kelly' via OxCal <ox...@googlegroups.com> wrote:
>
> Christopher?
> I’m guessing that 2000 would be “too many”?
> Bob
>
> Get Outlook for iOS

> To view this discussion visit https://groups.google.com/d/msgid/oxcal/MWHPR05MB34863FB831421FB010F9186DDDA8A%40MWHPR05MB3486.namprd05.prod.outlook.com.