You are here: Sampling > Sampling Methods > Fixed-interval Sampling

Fixed-interval Sampling

Fixed-interval sampling is generally the preferred sampling method because it produces the most consistent results. In fixed-interval sampling, you specify a selection interval i and one random start number r. The random start number must be greater than zero and less than or equal to the selection interval. The data is treated as a stream and the r-th item is the first item to be selected. Thereafter, every i-th item is selected.

An item is either a record or a dollar, so the 723rd item could be record 723 or the 723rd dollar in the population. For example, if the selection interval is 1,100 and the random start is 723, item 723 is selected first, followed by items 1,823, 2,923, 4,023 and so on.

Note: When you apply the fixed interval sampling method to a monetary unit sample, you are selecting individual dollars from the population. Therefore, a thousand dollar item contains 1,000 individual dollars and a million dollar item contains 1,000,000 individual dollars, any of which might be selected. Once an individual dollar is selected, the record containing this dollar is drawn.

With top stratum cutoff, applicable only to monetary unit samples, an item greater than the cutoff is automatically selected without affecting the sampling process. See Top Stratum Cutoff.

When you use fixed-interval sampling, you must be conscious of patterns in the data. Because a fixed interval is used for sample selection, a decidedly nonrepresentative sample can be drawn if the data has a pattern that coincides with the interval you specify. For example, you sample expenses with an interval of $1,000,000 and the monthly expenses that you test are also about $1,000,000. In this situation, it is possible that the same small expense category will be selected for every selection because this category appears at million dollar intervals in the file. However, such a situation is uncommon in practice.

Note: You must choose fixed interval sampling if you intend to use the Evaluate command to evaluate errors. The sample sizes determined by the Size command are also designed for use with fixed interval sampling.