Sampling Error
Sampling Error
Sampling error has occured if the sample statistics differ from the population statistics after the entire population has been examined. There are two types of sampling error: bias and dispersion. Bias or “lack of accuracy” occurs when the sample mean is different than the population mean. Bias can result from factors such as:
- Sampling only from the surface of a liquid at rest
- Sampling only from one edge of rolls or sheets
- Sampling from only one segment of the lot
- Instruments out of calibration
The result of a bias error is illustrated in Figure. Dispersion or “lack of precision” occurs when the measurements taken and recorded vary around the true measurement. Dispersion error is the result of variability in the sample standard deviation (see Figure). This type of error is typically due to improper reading or use of an instrument or an instrument that cannot read to the specified precision. The key to eliminating dispersion error is to choose the proper measuring instruments and make sure that the people using them are adequately trained in their use. The key to eliminating bias error is to make sure the instruments are calibrated and that the sample chosen accurately represents the population.
Summary
Remember that data provide the basis for all decisions and actions. It is important to know why data are needed and what they will be used for. Therefore, careful planning should precede every data collection effort. Be sure to collect only the data needed, and be sure that all information associated with the data and the data collection process is accurately recorded. When collecting the data, measure as accurately as possible within the given time and cost constraints. Also, all data collection should be done in such a way that the data can be easily used and understood.
Bias Error
Dispersion Error
Data collected from samples are used to make decisions. Therefore, it is critical that the sample be carefully chosen. When choosing a sampling scheme, several aspects should be considered: (1) the accuracy and reliability the scheme provides, (2) the additional cost in time and personnel that will be incurred, and (3) the timeliness with which the sample can be taken. If sampling is done properly, the data will accurately represent the population, and correct decisions can be made and appropriate actions taken. If sampling is not done properly, the data will not truly represent the population and the decisions made and actions taken may be the wrong ones. Therefore, whatever sampling method is used, it should be carefully designed.