R_Combine Charcoal

[Rayfo...@aol.com]

Hello Christopher.

 

I'm trying to R-Combine a set of Charcoal dates, but get multiple warnings.  OxCal 4.1b2

 

I'm obviously doing it wrong, but what?

 

Input:

 

Plot()
 {
  Outlier_Model("Charcoal",Exp(1,-10,0),U(0,3),"t");
  R_Combine("")
  {
   R_Date("A", 3975, 40)
   {
    Outlier("a", 1);
   };
   R_Date("B", 3985, 45)
   {
    Outlier("b", 1);
   };
   R_Date("C", 3985, 55)
   {
    Outlier("c", 1);
   };
   R_Date("D", 3850, 55)
   {
    Outlier("d", 1);
   };
   R_Date("E", 3740, 60)
   {
    Outlier("e", 1);
   };
   R_Date("F", 3985, 85)
   {
    Outlier("f", 1);
   };
  };
 };

 

Output:

Warning! No outlier model specified - use Outlier_Model()
Warning! No outlier model specified - use Outlier_Model()
Warning! No outlier model specified - use Outlier_Model()
Warning! No outlier model specified - use Outlier_Model()
Warning! No outlier model specified - use Outlier_Model()
Warning! No outlier model specified - use Outlier_Model()
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! ZERO distribution
Warning! Poor agreement - A= 44.0%(A'c= 60.0%)
Warning! Poor agreement - A= 44.0%(A'c= 60.0%)

regards

 

Ray Kidd

[Christopher Ramsey]

Hello

There are a number of things here.

1. The label on the Outlier should refer to the model - and so should
be "Charcoal" rather than "a", "b", "c" ... etc.

2. You cannot use a t-type outlier model within R_Combine - as the
assumption of R_Combine is that all of the samples have the same
radiocarbon concentration. So in this case you need to use Combine if
they are all deposited at the same time.

3. Using outlier analysis without a more general model is difficult as
there is nothing to control the range of possible ages.

However - just changing the first two will work:

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

Combine("")
{
R_Date("A", 3975, 40)
{

Outlier("Charcoal", 1);

};
R_Date("B", 3985, 45)
{

Outlier("Charcoal", 1);

};
R_Date("C", 3985, 55)
{

Outlier("Charcoal", 1);

};
R_Date("D", 3850, 55)
{

Outlier("Charcoal", 1);

};
R_Date("E", 3740, 60)
{

Outlier("Charcoal", 1);

};
R_Date("F", 3985, 85)
{

Outlier("Charcoal", 1);
};
};
};

The only problem being that the distribution is truncated - as the
program was unable to constrain the possible range of dates. The
following is better with some arbitrarily wide limits chosen:

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

Sequence()
{
Date(-2800);

Combine("")
{
R_Date("A", 3975, 40)
{

Outlier("Charcoal", 1);

};
R_Date("B", 3985, 45)
{

Outlier("Charcoal", 1);

};
R_Date("C", 3985, 55)
{

Outlier("Charcoal", 1);

};
R_Date("D", 3850, 55)
{

Outlier("Charcoal", 1);

};
R_Date("E", 3740, 60)
{

Outlier("Charcoal", 1);

};
R_Date("F", 3985, 85)
{

Outlier("Charcoal", 1);
};
};
Date(-2000);
};
};

Christopher

[Rayfo...@aol.com]

Hello,

 

Thank you for that, it is all much clearer now.

 

I was trying to see if I could identify the most probable outliers of charcoal only determinations.  By having the Charcoal outliers at probability (1) the outlier report has them all red and 100 %  (to be expected). 

 

However using your suggested Combine model I tried various other Charcoal probabilities, e.g. 0.05, 0.2 and 0.5.   This allowed the different posterior outlier probabilities to be seen more clearly on the green/red scale.  In particular it showed the latest two dates as least likely to be outliers and the four earlier more likely to be outliers.

 

It this a misuse of the function?

 

It also ties in with the Dsplit identification of two groups of four and two.  Presumably the two later determinations could date the context closest and indicate the earlier charcoal to be residuals?

 

Element	Ok	Outlier	Prior	Posterior	Model	Type
A			50.0	67.0	Charcoal	t
B			50.0	68.0	Charcoal	t
C			50.0	63.0	Charcoal	t
D			50.0	35.0	Charcoal	t
E			50.0	32.0	Charcoal	t
F			50.0	55.0	Charcoal	t

 

Plot()
 {
  Outlier_Model("Charcoal",Exp(1,-10,0),U(0,3),"t");
  Sequence()
  {
   Date(-2800);
   Combine("")
   {
    R_Date("A", 3975, 40)
    {

     Outlier("Charcoal", 0.5);

    };
    R_Date("B", 3985, 45)
    {

     Outlier("Charcoal", 0.5);

    };
    R_Date("C", 3985, 55)
    {

     Outlier("Charcoal", 0.5);

    };
    R_Date("D", 3850, 55)
    {

     Outlier("Charcoal", 0.5);

    };
    R_Date("E", 3740, 60)
    {

     Outlier("Charcoal", 0.5);

    };
    R_Date("F", 3985, 85)
    {

     Outlier("Charcoal", 0.5);
    };
   };
   Date(-2000);
  };
 };

 

 

 

regards

 

Ray Kidd

[Christopher Ramsey]

Well - I don't think this is what is really intended for the model.
In a sense what you are saying is that we have a 50% chance that each
of the samples is charcoal and a 50% chance it is something else - the
model is then selecting those samples that are more or less likely to
be charcoal. This identifies the samples that are older - but then
you can do this from the dates anyway.

Christopher

On 10 Nov 2008, at 17:31, Rayfo...@aol.com wrote:

> Hello,
>
> Thank you for that, it is all much clearer now.
>
> I was trying to see if I could identify the most probable outliers
> of charcoal only determinations. By having the Charcoal outliers at
> probability (1) the outlier report has them all red and 100 % (to
> be expected).
>
> However using your suggested Combine model I tried various other
> Charcoal probabilities, e.g. 0.05, 0.2 and 0.5. This allowed the
> different posterior outlier probabilities to be seen more clearly on
> the green/red scale. In particular it showed the latest two dates
> as least likely to be outliers and the four earlier more likely to
> be outliers.
>
> It this a misuse of the function?
>
> It also ties in with the Dsplit identification of two groups of four
> and two. Presumably the two later determinations could date the
> context closest and indicate the earlier charcoal to be residuals?
>