Interval/Difference yield impossible spans? Boundaries wrong?

[Jonah B.]

Hello everyone, 

I was trying to get the spans of usage for different burial sites. At first I used the Interval function which worked in a prior model but came back empty in this case. I played around with the syntax, having it within the sequence, within the phase, etc. 

Since I could not make it work, I decided to try the Difference function. This yielded some results however I noticed that the boundaries were off too. 

I have a burial site whose raw dates were all within the 4th millenium and the end boundary calculated as "from -1558 to -573".

My model for the burial sites without intervals or differences looks like this:

Plot()
{
 Outlier_Model("General", T(5), U(0,4), "t");

 Sequence("Site 1")
 {
  Boundary("Start Site 1");
  KDE_Plot("Site 1")
  {
   R_Date("MAMS 18914",3942,25){ Outlier("General", 0.05); };
   R_Date("MAMS 18915",3863,25){ Outlier("General", 0.05); };

  };
  Boundary("End Site 1");
 };

 Sequence("Site 2")
 {
  Boundary("Start Site 2");
  KDE_Plot("Site 2")
  {
   R_Date("MAMS 23721",3870,26){ Outlier("General", 0.05); };
   R_Date("MAMS 23720",3775,25){ Outlier("General", 0.05); };
  };
  Boundary("End Site 2");
 };
};

Why are the boundaries messed up and how can I fix that?

I'd appreciate any help! Thanks, Jonah

[MILLARD, ANDREW R.]

Hello Jonah,

I don' get that result with your model but just pasting and running I get this:

Name		Unmodelled (BC/AD)		Modelled (BC/AD)		Indices Amodel=94.5 Aoverall=94.9					Controls
		from_95.4	to_95.4	from_95.4	to_95.4	Acomb	A	L	P	C	Select	Page	Edit
Outlier_Model General				-216	187					99.8	20
T(5)		-2.65	2.65							98.6	2
U(0,4)		3.98986e-17	4	5.37764e-17	3.948		100			99.9	3
Sequence Site 1											4
Boundary Start Site 1				-3244	-2346					99.7	5
KDE_Plot Site 1										99.8	6
N Site 1_Kernel		-2	2	-2.79	1.99		96.6			4.9	7
U Site 1_Scale		9.97466e-18	1	0.04	1		100			9.5	8
R_Date MAMS 18914		-2566	-2310	-2561	-2306		96.5		96.6	99.9	9
R_Date MAMS 18915		-2458	-2209	-2463	-2233		100.7		96.6	100	10
Boundary End Site 1				-2456	-1439					99.6	11
Sequence Site 2											12
Boundary Start Site 2				-3204	-2214					99.7	13
KDE_Plot Site 2										100	14
N Site 2_Kernel		-2	2	-2	2		100			99.9	15
U Site 2_Scale		9.97466e-18	1	0.055	1		100			100	16
R_Date MAMS 23721		-2462	-2209	-2457	-2206		94.4		96.3	99.9	17
R_Date MAMS 23720		-2290	-2063	-2296	-2066		96.5		96.3	99.9	18
Boundary End Site 2				-2287	-1339					99.5	19

There are problems with convergence as can be seen in the plot where the boundaries do not have decreasing tails far from the dates and are truncated. 

Best wishes 

Andrew 

-- 

Prof. Andrew Millard 

Department of Archaeology, Durham University, UK 

Email: A.R.M...@durham.ac.uk  

Personal page | Dunbar 1650 MOOC 

StratoBayes: De-risking subsurface decision making

 

---

From: ox...@googlegroups.com <ox...@googlegroups.com> on behalf of Jonah B. <jonah.b...@gmail.com>
Sent: 07 August 2026 10:49
To: OxCal <ox...@googlegroups.com>
Subject: Interval/Difference yield impossible spans? Boundaries wrong?

 

You don't often get email from jonah.b...@gmail.com. Learn why this is important

[EXTERNAL EMAIL]

--
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/b9bca968-b7fa-45eb-9819-8b1682e147een%40googlegroups.com.

[Jonah B.]

Hello Andrew, 

that's odd, I will try again then...

thank you!

[Christopher Ramsey]

Single phases with only two events are really over-parameterised hence the very long tails. It is not really possible to deduce the length of a phase from only two representative events.

Best wishes

Christopher

> On 7 Aug 2026, at 11:22, Jonah B. <jonah.b...@gmail.com> wrote:
>
> Hello Andrew,
> that's odd, I will try again then...
>
> thank you!
>
> MILLARD, ANDREW R. schrieb am Freitag, 7. August 2026 um 12:15:38 UTC+2:
> Hello Jonah,
>
> I don' get that result with your model but just pasting and running I get this:
>

> NameUnmodelled (BC/AD) Modelled (BC/AD) Indices

> Amodel=94.5
> Aoverall=94.9 Controls
> from_95.4 to_95.4 from_95.4 to_95.4 Acomb A L P C Select Page Edit
> Outlier_Model General-216 18799.8 20
> T(5)-2.65 2.6598.6 2
> U(0,4)3.98986e-17 4 5.37764e-17 3.94810099.9 3
> Sequence Site 14
> Boundary Start Site 1-3244 -234699.7 5
> KDE_Plot Site 199.8 6
> N Site 1_Kernel-2 2 -2.79 1.9996.64.9 7
> U Site 1_Scale9.97466e-18 1 0.04 11009.5 8
> R_Date MAMS 18914-2566 -2310 -2561 -230696.596.6 99.9 9

> R_Date MAMS 18915-2458 -2209 -2463 -2233100.796.6 100 10

> Boundary End Site 1-2456 -143999.6 11
> Sequence Site 212
> Boundary Start Site 2-3204 -221499.7 13
> KDE_Plot Site 2100 14
> N Site 2_Kernel-2 2 -2 210099.9 15
> U Site 2_Scale9.97466e-18 1 0.055 1100100 16
> R_Date MAMS 23721-2462 -2209 -2457 -220694.496.3 99.9 17

> R_Date MAMS 23720-2290 -2063 -2296 -206696.596.3 99.9 18
> Boundary End Site 2-2287 -133999.5 19

>
>
> There are problems with convergence as can be seen in the plot where the boundaries do not have decreasing tails far from the dates and are truncated.
>
>

> Best wishes
> Andrew
> --
> Prof. Andrew Millard
> Department of Archaeology, Durham University, UK
> Email: A.R.M...@durham.ac.uk Personal page | Dunbar 1650 MOOC StratoBayes: De-risking subsurface decision making

> To view this discussion visit https://groups.google.com/d/msgid/oxcal/3e60bf1f-912b-4da3-8a57-bc4f1f55d490n%40googlegroups.com.

[Jonah B.]

In my model I had much more, I just omitted those for simplicity but I'm seeing that this might have been unproductive...