Hard to understand low agreement for a date in a model

[Patricio De Souza]

Hello, everyone. I’d like to ask for help with the model below. This is a model based on six charcoal dates extracted from an archaeological site. Of these six dates, two are in stratigraphic relationship in one column (“195A”), and another two are also in stratigraphic relationship in another column (“195B”). The remaining dates do not have a clear stratigraphic relationship with the others.

My problem is that the date Ranl-195-2c consistently gives me a very low agreement (ca. 31%). I can’t figure out what’s going on, since this date is clearly older than the one that follows it in the “195B” sequence.

Thanks in advance for any help. Best regards.

Plot()
 {
  Curve("SHCal20","shcal20.14c");
  Sequence (Ranl-195)
  {
   Boundary ("Start Ranl-195");
   Phase ("Ranl-195")
   {Sequence ("195A")
    {
     Boundary ("Start 195A");
     R_Date("Ranl-195-2a",5883,35);
     R_Date("Ranl-195-2b",5921,32);
     Boundary ("End 195A");
    };
    Sequence ("195B")
    {
     Boundary ("Start 195B");
     R_Date("Ranl-195-2c",6050,29);
     R_Date("Ranl-195-2d",5861,33);
     Boundary ("End 195B");
    };
    R_Date("Ranl-195-2e",5913,34);
    R_Date("Ranl-195-2f",5962,32);
   
   };
   Boundary ("End Ranl-195");
   Span("Ranl-195");
  };
 };

[Erik Marsh]

Hi Patricio – 

I think the problem is too many boundaries (you have six dates and six boundaries), which can over-constrain a model with few dates. The two overall site boundaries are enough here, especially if the occupation is short. When I remove the other four, the A for date 2c goes up to 60, which is okay if you're sure of the model's assumptions. It's clearly the earliest date in the phase, so it's worth checking to see if this represents a separate, earlier occupation, or is perhaps simply a statistical outlier. As an example (below), I've assumed (without knowing!) that this sample is charcoal add tagged it as an outlier. This down-weights its influence on the model, the A goes up to 89, and its distribution falls in line with the rest of the phase. You could also use a General outlier model; it depends on the sample and stratigraphy. Doing this narrows the likely span of occupation to about two generations – I have no idea if this is feasible or not based on these contexts and the rest of the material at the site.

A few other tweaks: I moved Span to within the phase, so it gives you the time between the site boundaries. I replaced Phase with KDE_Plot, which has the same function in the model, but also produces a KDE of the site's occupation. This is usually a useful summary of the modeled dates. (In a published version, it's best practice to include lab codes in the names of each date in the model).

Hope this helps

Erik

  Plot()

 {
  Curve("SHCal20","shcal20.14c");

  Outlier_Model("Charcoal",Exp(1,-10,0),U(0,3),"t");

  Sequence (Ranl-195)
  {
   Boundary ("Start Ranl-195");

   KDE_Plot("Ranl-195")
   {
    Sequence ("195A")
    {

     R_Date("Ranl-195-2a",5883,35);
     R_Date("Ranl-195-2b",5921,32);

    };
    Sequence ("195B")
    {
     R_Date("Ranl-195-2c",6050,29)
     {
//Assuming this sample is charcoal
      Outlier("Charcoal",1);
     };
     R_Date("Ranl-195-2d",5861,33);

    };
    R_Date("Ranl-195-2e",5913,34);
    R_Date("Ranl-195-2f",5962,32);

    Span("Span Ranl-195");

   };
   Boundary ("End Ranl-195");

  };
 };

[Patricio de Souza]

Hi Erick. You're absolutely right; the problem was the excessive and unnecessary number of boundaries. Thanks also for the other tips, such as using KDE_Plot and changing the position of the Span function. The sample is indeed from charcoal, but I think it might be unnecessary to apply the outlier model, since the A values for the date and the model seem sufficient without applying the outlier model for charcoal.

Thanks again; your advice will also be useful for other models I’m working on.

Best regards,

Patricio

--
You received this message because you are subscribed to a topic in the Google Groups "OxCal" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/oxcal/AQSZuLlrr10/unsubscribe.
To unsubscribe from this group and all its topics, send an email to oxcal+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/oxcal/54d94c5b-3618-4ab1-b506-3fc9812be792n%40googlegroups.com.