In a measurement process sampling is extremely important. All subsequent steps depend on it. If the sample is inadequate, confidence in the measurement result and the resulting decisions will be compromised. Due to the importance of the samples, ABNT NBR ISO / IEC 17025: 2017 establishes, in subsection 7.3, general requirements for them to be obtained in an appropriate way. The laboratory must have a sampling plan, when sampling substances, materials or products for testing or calibration. The sampling method must address the factors to be controlled, to ensure the validity of the measurement results. Sampling plans should, whenever reasonable, be based on appropriate statistical methods in order to allow two important characteristics of the sample to be preserved: impartiality and representativeness. Therefore, among the aspects to be considered, the following can be highlighted:
Nature of the substances, materials or shape of the object of interest in a study;
Sample size;
The location of the object of study;
Position of obtaining the sample;
Frequency of obtaining the samples;
Sample collection procedure;
Cost (materials, reagents, equipment, personnel, among others);
Training of the collector (depends on the complexity involved in obtaining the sample).
In a way, all of these aspects are closely related. For example, a very large sample increases sampling costs. The sample's position or location has an influence on the sample's representativeness. Collector training, among other aspects, can compromise impartiality, resulting in a bias in the results. It should be noted, however, that there are cases where some of these aspects are already duly defined in standards or measurement method and must be consulted and followed to allow comparability of results.
The SAS Calc is designed to establish the minimum sample size in order to estimate the arithmetic mean of a specific property, considering the following aspects:
Metrological quantity to be determined consists of a continuous random variable (for example, the length of an object);
Data population is accepted as infinite compared to the sample size;
Population standard deviation is unknown;
The distribution of population data appears as a Normal distribution (Gaussian probability distribution);
Simple random sample.
To test the normality of the pilot sample, the ACIC NORMALIDADE SOFTWARE can be used. It is available at the following adress:
https://sites.google.com/site/acicnormalidade