Sampling techniques

Sampling techniques

Random numbers

Sampling may be required when carrying out root cause analysis. Although the techniques applied would be more relevant on an industrial scale, awareness of the different techniques can be useful.


Random

Random numbers (e.g. 7, 54, 29, 2) can be used to determine units to be drawn from a larger population. Tables, dice or computers can be used to generate the numbers.
Example: Use if wishing to select random dates for records to examine over a month e.g. 29, 7, 30, 19, 8 and 15. Note the maximum upper level must be stated, in this case 28, 29, 30 or 31.


Systematic

Measures at fixed intervals of time, numbers,length etc.
Example: Can use if wishing to observe temperature records at specific times in a chiller cabinet over several days, e.g. 00:43, 03:33, 06:54, 14.01, 03:18 and 22:20.


Stratified

This tool should be used if there are known differences between categories.
Example: There are concerns about a supplier that supplies a family chain of restaurants. The restaurants are the same in the majority of aspects except Restaurant A serves 150 people a week where as restaurants B and C serve only 50 people. It may be chosen to look at 3 times as many records from restaurant A compared with the amount of records examined at restaurants B and C.

Cluster

When the population is known to be stable without much variation. A group of units taken to represent the whole population.
Example: There are 400 records to look at with not much variation in data. Examine 40 results and attribute findings to be indicative of what would be found if all 400 records were examined.