Ranks is a feature parameter which can be used to place a limit upon the number of ranks (categories or classes) to which a feature can be assigned. Ranks can be specified in the Features table. Right clicking on the column brings up a popup menu that allows you to set all features with one click.
If the value of Ranks is 0, there is no limit to the number of ranks or classes the data might be assigned to. For example, data values might represent the number of stars that appear on a flag and in theory there is no limit to the number. However, all flags with six stars would belong to one rank and all stars with seven would belong to another. Discrete seriation (which uses Manhattan distance and Ranked data) would try to order artifacts (flags) so that flags with six stars are in closer temporal proximity to flags with seven stars then they are to flags with eight stars. Nominal seriation (which uses Hamming distance and Classed data) would only try to keep like flags together but not order them by their categories, the numbers of stars.
If the value of Ranks is equal to 1, any data (including alphanumeric) can be converted automatically to ranked data. If the value of Ranks is 1, it is assumed that there is only one possible rank or class and the data is taken to represent presence or absence of, or membership or non-membership in, that class. Any non-zero entry in the data is taken to be a member of the class (present). Zeroes and blank entries are interpreted according to the feature's Zeroes and Blanks parameters respectively (values of "Absent & Unknown" and "Absent & Zero" are taken to be not members of the class, values of "Present & Unknown" and "Pressent & Zero" are taken to be members of the class). The Ranks parameter allows you to easily apply occurrence seriation even when the data were recorded in something other than the standard 0/1 format. You may have used numbers to represent the number of occurrences, or a measurement, of a feature, and blanks to represent absence of a feature; or you may have used descriptive terms like "smooth", "serrate" and "dentate" to describe your artifacts. To use occurrence seriation there is no need to convert your data to zeroes and ones. Instead you can set the value of Data to Classed, Ranks to 1 and Blanks to Absent & Zero and OptiPath will implicitly convert the data for you (you can see the result in the Values table).
A Ranks value of 2 is useful in applying occurrence seriation to numerical Ranked data that represents measurements over a wide range of values. For example, if data ranges from 3 to 48 for a feature, a Ranks value of 2 will result in all values from 3 to 25 being assigned rank 0 and values from 26 to 48 being assigned rank 1. This is equivalent to dividing the artifacts into "big" and "small" and using occurrence seriation. Similarly using a Ranks value of 3 is equivalent to dividing artifacts into "big", "medium" and "small". By setting the Ranks to n with Ranked data, you can separate your numerical data into n classes or categories and seriate them so artifacts with similar data values will tend to be more temporally proximate than artifacts with very different values.
If the value of Ranks is greater than 0, Measured data can be converted automatically to ranked data. If the value of Ranks is n, an integer greater than 1, then the data values for a class are divided evenly into n classes. The Ranks parameter can be used to consolidate a wide range of numerical data into a smaller number of categories. For example, raw numerical data including fractional values can be converted into two or three categories representing "present" and "absent" or "large", or "medium" and "small". The categories will be assigned integral values and OptiPath will attempt to preserve their numerical ordering in any seriation. If there is a zero in your data (or a blank which is to be interpreted as zero) the ranks will be numbered so that zero will be in the rank numbered 0. For example, if your data ranged from -10 to 4, with five ranks, the ranks would range from -3 to +1 and values from -10 to -8 would be assigned to rank -3, values from -7 to -5 would be assigned to rank -2, values from -4 to -2 would be assigned to rank -1, values from -1 to +1 would be assigned to rank 0, and values from 2 to 4 would be assigned to rank 1. Discrete seriation generally uses a Ranks value of n greater than 1.
Measured data must have a Ranks value of 0. Classed data must have a Ranks value of 0 or 1. Ranked data can have a Ranks of value 0, 1 or any integer greater than one.
To see the effect on the raw data of setting various feature parameters, including the Ranks parameter, look at processed data in the Values table.