[Up][Contents][Index]
Information from Analysis
In the first instance this program is designed to take into account
stratigraphic information from a site and modify the probability
distributions obtained directly from radiocarbon calibration or other
dating methods (called `prior' probability distributions) in the
light of this additional data (producing so called `posterior'
probability distributions).
There are however other, equally important, types of information
which can be obtained.
See also [Program Operation]
For a given group of dates which may be constrained in some way
by stratigraphic information it is useful to be able to obtain a
probability distribution for the first and last members of the group.
eg:
Sequence
{
Boundary Start;
Sequence
{
R_Date 800 35;
Phase
{
First;
R_Date 750 30;
R_Date 800 35;
Last;
};
R_Date 660 35;
};
Boundary End;
};
It should be stressed that to use this to estimate the start and end
of phases relies on the fact that the distribution of dated samples
within the group are representative of the archaeological phase in
question.
If no objects have been recovered from the first century of a period
no amount of statistical analysis can determine when that period
began!
Furthermore if there are no dated events prior to the period and a
large number of dated events within it statistical analysis is
liable to indicate that the period started earlier than it
actually did simply because of the inevitable scatter in the
measurements.
These caveats are no more or less relevant to non-mathematical
methods of analysis and simply imply good archaeological practice
in bracketing periods.
See also [Using boundaries]
[Program Operation]
[Example]
[Mathematical Methods]
The next type of information which one might wish to glean from the
analysis is the span of a group of dates.
A probability distribution can be generated which represents the
difference in age between the first and last items in a group. eg.:
Sequence
{
Boundary Start;
Sequence
{
R_Date 800 35;
Phase
{
R_Date 750 30;
R_Date 800 35;
Span;
};
R_Date 660 35;
};
Boundary End;
};
Clearly you should bear in mind the caveats mentioned in the
preceding section.
See also [Program Operation]
[Example]
[Mathematical Methods]
The two previous sections outline one way in which a group of dated
events can be treated in relation to archaeological phases.
This approach assumes that the dated events are both well constrained
and cover the archaeological phase from start to finish.
An alternative approach is to assume that the deposition of dated
artifacts is fairly uniform chronologically and use the distribution to
estimate the boundaries of the archaeological phases using this model.
This is the other function of the
Boundary statement used to
mark which samples come from a set period. eg:
Sequence
{
Boundary Start;
Phase
{
R_Date 750 30;
R_Date 830 30;
R_Date 820 30;
R_Date 760 30;
R_Date 810 30;
R_Date 800 30;
};
Boundary End;
Span;
};
Using such a model will give a much more realistic estimate of the
phase boundaries than simply assuming that the events are
unconstrained (ie not using Boundaries at all) and estimating when the
first and last events took place.
If the phase is well constrained anyway the results will be
very similar.
See also [Program Operation]
[Example]
[Mathematical Methods]
It is often useful to be able to find out what the interval between
two phases or two events was.
A probability distribution can be obtained for such events which
follow one after the other in a sequence. For example the fragment:
Sequence
{
R_Date 800 35;
Interval;
R_Date 660 35;
};
It is also possible to calculate a probability distribution for the
difference between any two events in an analysis
(Difference).
See also [Program Operation]
[Example]
[Mathematical Methods]
Sometimes you may wish to estimate the probability of various
possible orders of dated events.
Assuming that the dating evidence is good enough to provide the
necessary discrimination such probabilities can easily be calculated
(Order).
See also [Program Operation]
[Example]
[Mathematical Methods]
Clearly any analysis relies very strongly on the reliability of the
information included.
The analysis does include the calculation of some overall indicators
of how well all of the data incorporated in the analysis agrees and
which elements of the data are most suspect.
It is frequently the case that there is some uncertainty associated
with the stratigraphic evidence for an item (or indeed the date
measurement itself). In these cases it is necessary to be able to
find out how likely an item is to be in a particular place in a
chronological sequence.
If the position of an item is questioned
(Question)
in this way the item will
be ignored in the main analysis and a probability calculated.
Consider for example the fragment:
Sequence
{
R_Date 970 35;
R_Date 1180 30?
R_Date 930 35;
};
This would give a fairly low probability of being true
(in fact 0.7%).
No provision has been made for assigning probabilities to the
veracity of dated events as such a practice seems rather arbitrary
and virtually impossible to justify.
See also [Program Operation]
[Mathematical Methods]
The resultant probability distributions after analysis are not
in general independent.
For example two events in a sequence may have probability
distributions which overlap but clearly given the fact that
they are in a sequence the second one must always follow the first.
It is, therefore, occasionally useful to be able to display a
plot of one distribution relative to another.
This is can be achieved
(Correlation)
although it should be said that the
resultant two dimensional map needs some practice in interpretation.
See also [Program Operation]
[Example]