Seriation (archaeology)

The term seriation is derived from the word series, which designates a specific set of similar things and consequences. In this sense it is also a design principle to achieve through regular repetition of free or predetermined forms a perfect symmetry, eg the meander.

In archeology, a method for sparse matrices order is named after the unimodal model with seriation. It is used for relative dating of artifacts from various sites in order to bring them into a chronological order.

This " sparse matrices " are usually larger tables that list the closed findings on a specific topic in the lines ( eg waste pits, graves ) and in the columns specific fund types ( pottery, jewelry, weapons). Reported the tables either the occurrence and non- occurrence of these types ( "zero" or "one", also: Anwesenheits-/Abwesenheits-Matrix ) or the frequency of occurrence of the types (frequency matrix).

In everyday life, we encounter many phenomena which are subject to the linear model: The more gasoline fueled, the higher of the bill. A unimodal model exists when a phenomenon often initially, after a maximum, however, again often. This assumption make archaeologists phenomena along the time axis: something does not yet exist; it is invented and then dive occasionally; it is becoming increasingly popular and appeared frequently on; it will be replaced by something new again unfashionable and less frequently in the Fund spectra to disappear. The seriation is the adequate mathematical procedures, tables, subject to such phenomena, suitable to organize so that they show in the result, the closed finds and the types in a chronological order.

The procedure was according to different - introduced to order a presence / absence matrix precursors by Klaus Goldmann 1972 - slightly different nature. Then it was steadily improved, including in the 1980s by Peter Him by taking into account the frequencies. Further research led to the discovery that the optimal type of diagonalization can be achieved by means of a correspondence analysis; the first solution to this potentially multi-dimensional process is identical to the result of a seriation.