Seriations
Seriations are orderings of items or objects, or in our case, of artifacts or assemblages. OptiPath has been developed as a tool for seriating archaeological artifacts. The object of seriation is to determine an ordering of a collection of artifacts that replicates as closely as possible the historical events they represent. To do this, OptiPath relies on the assumption that the artifacts share characteristics or features whose measure evolves gradually over time. The ordering of artifacts that produces the most gradual evolution of all features for all artifacts is considered the shortest path seriation. Alternatively, one could optimize unimodality to find a best seriation.
In OptiPath, a file refers to a permanent file retained on your computer which holds all of the data for all your data sets and seriations. A data set refers to a single set, or sample, of archaeological artifacts which you are trying to seriate. You may have many data sets (for example, separate data sets for Hohokam pottery, Hawai'ian fish hooks and Inuit spear points), but they can all reside in a single file. A seriation refers to a single attempt to seriate a data set. You may perform, and save, several separate seriations for a single data set. There are different techniques of seriation and you may try more than one to find the best possible seriation (as reflected by the most convincing resulting evolution of the various features of your data).
Seriations Menu
Show/Hide Table allows you to open/close the Seriations table and create, delete, select and edit the seriations, and edit their parameters.
Show/Hide Results is a toggle that allows you to display seriation results in the Seriations table.
Show/Hide Errors is a toggle that allows you to display seriation error measurements in the Seriations table.
Help will bring up this documentation.
Seriations Table
By choosing Hide Results and Hide Errors in the Seriations menu, you can reduce the number of fields displayed.
In the Seriations table you may add, edit and delete seriations. Only the active data set can be edited, other rows are greyed out. Right clicking on the table will bring up a pop up menu of available actions. There are a number of attributes associated with each seriation:
Active indicates the active seriation. OptiPath allows you to create a number of seriations for the same data set. The Active seriation indicates the current seriation, or the one you are working with now. The data you look at in the various tables is always from the active seriation. You may reload an old seriation to examine its inputs and results by setting it to be the active seriation. When you create a new seriation it is automatically the active seriation. To deactivate the current seriation you can reactivate an old one or create a new one.
Seriation a name that uniquely identifies the seriation for the current data set. Seriation names are restricted to 50 characters. The seriation name is required for each seriation. Each time you create a new seriation, it will be given a default name of the data set name plus the date and time the seriation is created. You may change the name of a seriation at any time, as long as you do not try to give it the same name as another seriation in the same data set.
Description is an optional description that can be entered for each seriation. There is no limit to the length of a description.
Author an optional field where you can record the name of the person who is performing the seriation. This is useful for historical purposes when you come back and look at a large number of seriations in your file and you wonder who did what.
Created an information field which gives the date the seriation was created. This is useful for historical purposes when you come back and look at a large number of seriations in your file and are looking for a seriation done at a particular time. This field is read only; it cannot be modified.
Gradual Index is a measure of how gradual a seriation is. See Objectives for more information.
Unimodal Index is a measure of how unimodal a seriation is. See Objectives for more information.
Path Length is the length of the path determined by the seriation. See Objectives for more information.
Avg Rate is the normalized average rate of change of the features of artifacts from one item to the next, assuming the total path length is 1. See Objectives for more information.
Max Rate is the normalized maximum rate of change of the features of artifacts from one item to the next, assuming the total path length is 1. See Objectives for more information.
Sqd Rate is the sum of the normalized squared rates of change of the features of artifacts from one item to the next, assuming the total path length is 1. See Objectives for more information.
Date Error is the average error in the seriation for items with user supplied dates. If the user enters an Earliest and/or Latest date for an item (an artifact or assemblage), OptiPath compares the date of an item computed by OptiPath with the user supplied dates.
Order Error is the average order error in the seriation. To compute errors, OptiPath compares the Index of an item with the Order computed by OptiPath.
% Date Error is the average percentage date error in the seriation for items with user supplied dates. If the user enters an Earliest and/or Latest date for an item (an artifact or assemblage), OptiPath compares the date of an item computed by OptiPath with the user supplied dates and express the difference as a percentage of the allowed range of dates in the seriation (determined by the difference between the Earliest and Latest dates for the seriation).
% Order Error is the average percentage order error in the seriation. To compute errors, OptiPath compares the Index of an item with the Order computed by OptiPath and express the difference as a percentage of the range of ordinal positions in the seriation, which is just the number of items included in the seriation.
Technique indicates the seriation technique OptiPath should use. OptiPath can use a number of different seriation techniques, all using OptiPath's heuristic optimization algorithms which try to find the best possible seriation conforming with the technique and the parameter settings you have chosen. The seriation technique options include Custom, Shortest Path, Frequency, Occurrence, Nominal and Discrete. By choosing a seriation technique (other than Custom) you are selecting a set of default feature parameter settings which are applied to all features and which you will not be able to change. The Custom technique allows you to choose your own feature parameter settings for each feature.
Objective allows the user to choose the objective to be used in determing a best possible seriation. See Objectives for more information.
Unimodal Weight allows the user to set the weights for an objective function that is a weighted combination of gradual and unimodal. The unimodal weight must be between 0 and 1. In seriating OptiPath will weight the unimodal by the unimodal index and the gradual index by 1 - the unimodal index. See Objectives for more information.
Break Ties directs the algorithm to break ties between alternative solutions having the same objective value. If the user is using Unimodality as the Objective, ties will be broken using the Path Length. If the Objective is Path Length, or Average, Maximum or Squared Rates of Change, then Unimodality is used to break ties.
Earliest is the earliest allowed date for any item in a seriation. If you do not set a date OptiPath will use 0 by default.
Latest is the latest allowed date for any item in a seriation. If you do not set a date OptiPath will use 1000 by default.
Use Dates indicates whether OptiPath should be constrained by the Earliest and Latest dates in the Artifacts (Assemblages) table. While seriating, OptiPath will only consider an individual artifact/assemblage's Earliest and Latest dates if Use Dates has also been selected for that item in the Artifacts (Assemblages) table. If Use Dates is not selected here but the individual artifacts/assemblages do have dates entered for Earliest and Latest, then OptiPath will seriate ignoring the user's Earliest and Latest dates and then the Dates Graph and Date Error will indicate how well OptiPath's dates align with the user's. When Use Dates is not checked on the Seriations table, Use Dates on the Artifacts/Assemblages table also tells OptiPath which dates to use when computing date Errors.
Use Frequencies indicates whether OptiPath should use the data as input or to compute frequencies from the data and use the frequencies in seriation. If Use Frequencies is checked and Normalize is also checked for any feature in the Features table (which we suggest you think out very carefully before doing), then OptiPath will compute the frequencies first and normalize the values second. If frequency seriation has been chosen as the seriation technique Use Frequencies will be chosen automatically and the user will not be able to modify the choice until a different seriation technique is chosen.
Use Assemblages indicates whether or not you want OptiPath to aggregate artifacts into assemblages and seriate the assemblages rather than the artifacts themselves.
Use Weights indicates whether or not you want OptiPath to weight each feature according to the weights specified in the feature parameters. If Weights is not selected, a weight of 1 is used for each feature rather than any values set in the feature parameters.
Randomize directs OptiPath when to generate solutions somewhat randomly when using OptiPath's heuristic solution techniques. This may sound crazy but the controlled use of randomness can actually improve the performance of many heuristic techniques. The drawback is that it is not reproducible (such is the nature of randomness). If you want to be able to replicate your results at some future date, then do not use Randomize. But if you want to see if OptiPath can find a little better answer, you might try Randomize. If you want the best of both worlds, do not use Randomize but set Seed (see below).
Seed if you want the advantages of the Randomize feature (see above) but want to be able to replicate your results, you may enter a seed value (any positive integer). Essentially, the seed is the starting point for a random number generator. Selecting a seed value is equivalent to running OptiPath with the Randomize option but you determine the randomly selected starting point, rather than the computer. This way you have replicable random results. As long as you set the same seed value, you should get the same results; but if you change seeds, you could get a different answer (but not necessarily). If Randomize is selected, OptiPath generates its own seed randomly (based on the time and date so it will not repeat itself on subsequent trials) and you may not be able to replicate its results. There is also an opportunity to randomly select a new seriation as a starting point for the algorithms (see Shuffle).