Top stratum cutoff relates only to monetary unit samples that use one of the two Interval sampling methods. If a monetary unit interval sample is selected, before sampling, Analyzer automatically stratifies the population into a “top stratum” and a “bottom stratum,” based on the absolute value of the field. Items equal to or exceeding the cutoff value are automatically selected and are not subject to the sampling algorithm.
The effect is as if the data were separated into two files; one file for each stratum. Items in the top stratum are selected immediately. The selection process is then applied to the bottom stratum as if the greater amounts were not in the file.
By default, the cutoff for a monetary unit interval sample is the size of the interval. This means that rather than merely having a high chance of selection, which in most cases equals 1, the item is guaranteed to be selected. If this is not what you want, enter a value in the Cutoff text box of the Sample dialog to separate the top and bottom strata.
To remove top stratum cutoff, specify an arbitrarily large cutoff value so that none of the items in the population will fall into the top stratum.
For example, if you were sampling with an interval of $1,000,000, all items with an absolute value equal to or greater than $1,000,000 would be in the top stratum and automatically selected. If, in the Sample dialog, you entered 500000 in the Cutoff text box, an item with a value of $500,000 would fall in the top stratum and be automatically selected, even though its chance of selection would have been only 50% using the sampling algorithm.
Similarly, if you specified 10000000 in the Cutoff text box, an item with a value of $5,000,000 would fall in the bottom stratum and have a proportionate chance of selection. While the probabilities may never exceed 1 because this item’s value is 5 times the interval, its chance of selection would nominally be 5, indicating that it will likely be selected 5 times.
Note: Modifications to the top stratum cutoff must be made only by specialists experienced in the technical theories relating to statistical sampling.