Variable Filtering - Setup
|Top Previous Next|
When variable filtering is selected, you are initially presented with an options screen which can be used to tune the annealing and also select key features of the filter options.
Iterations / temperature set – This option sets the maximum number of rearrangements of the species filter order that will be attempted at each temperature.
Successes per Temperature Step – This option determines how many superior arrangements in the species order will be found at each temperature before the temperature declines.
Starting temperature – The higher the initial temperature the more mixing of the species order will occur. In practice a temperature above 100 (boiling point) will be unlikely to be useful.
Proportional decrease in T per Step – This option should be less than 1. A value of 0.2 will result in a reduction to 20% of the initial value at each step.
Minimum shuffle size – This sets the minimum length of the species sequence that will be rearranged. For example if it is set at 3, no sequence of fewer than 3 species will be moved or reversed.
Minimum abundance – For presence/absence data this should be set to 0. For quantitative or percentage composition data it can be set to a higher number. If it is set, for instance, to 4, then negative filters will only retain sites (samples) where the abundance of the species exceeds 4.
Filter type - This allows a switch from negative to positive filtration. Negative is normally recommended. In negative filtration a sample is retained if the species is present.
While the computation is running, the graph shows the progress of the calculation downwards towards the most stable solution. To show the chart of the whole series of iterations, untick the box marked 'Track 350'. Tick the box to chart just the last 350 iterations. When the computation has finished, the progress chart disappears, and the results are shown on the Filters and Filter Plot pages. You can halt the computation at any time by pressing the 'End' button; the most recent iteration will be shown on the Filter Plot page: