Dealing with zeros or sparse data
|Top Previous Next|
Rows or columns in the working data holding zero values can be removed. On the Working Data page select Handling zeros in the Type of Adjustment panel.
Select the adjustment to be made and click on Submit to make the change. The adjustments are not made to the original data (on the Raw Data tab), but to the working data.
The possible options are as follows:
Deselect Rows: Removes rows not wanted in this analysis.
Deselect Columns:Removes columns not wanted in this analysis.
Delete zero columns - Every column in the data set that only contains zeros is removed.
Delete zero rows - Every row in the data set that only contains zeros is removed.
Remove sparse columns - Every column in the data set which contains < x non-zero elements is removed. The value of x is entered by the user in the At least x non-zero value text box.
Remove sparse rows - Every row in the data set which contains < x non-zero elements is removed. The value of x is entered by the user in the At least x non-zero value text box.
Beals - Beals smoothing is used on sparse data sets where many of the samples hold a small proportion of the total species list. As the transformation only uses the presence/absence of species, it should not be applied to good quality quantitative data as you will be discarding much of your information. It should only be performed on data arranged with the samples as columns. Described and discussed by Beals (1984) and McCune (1994) respectively, each element, eij, in the data array is replaced by:
where i is the column (sample) number, j is the row (species) number, Ni is the number of species in column (sample) i, Bothjk is the number of samples holding both species j and k, Samplesk is the number of samples holding species k and maxS is the total number of columns (samples).