A sample survey requires the study of small portions of a population, as there may be a certain amount of inaccuracy in the information collected during the sampling analysis. This inaccuracy is called the sampling error or error variance. On the other hand, non-sampling errors (systematic errors) mainly occur due to errors of computation at the state of classification and processing of data. Errors in sampling can be classified as random error and systematic error or sampling errors and error of measurement. Sampling error is a function of sampling size and systematic error is the result of non-sampling factors like study design, the correctness of execution sample frame errors, random sampling error, and non-response error. Random sampling error and systematic error associated with the sampling process may combine to yield a sample that is less than perfectly representative of the population.
These errors are not a measurement error not a systematic bias in the sample. It is the error that depends on the representativeness of the sample. The precision of the sample is greater when the sampling error is less. Sampling errors are also classified as biased errors and unbiased errors. The process of selection and estimation of samples may have some bias which leads to biased errors. Judgment sampling is used in a research survey instead of simple random sampling; some bias is introduced in the result due to the judgment of the investigator in selecting the sample. These errors are biased sampling errors. Unbiased errors are caused due to disagreement between the population units selected in the sample and those not selected. Errors occur in the final result due to the fact of the difference in the unit.
Bias may arise due to faulty selection of samples or substitution. The easiest method of avoiding bias is by selecting the sample randomly. Sampling errors are reduced by increasing the sample size.
Non-sampling errors mainly occur due to incorrect method of interviews, lack of experience of investigators, inadequate data specification, errors in data processing operations, errors in classification of data, etc.
These sampling errors can be reduced by controlling all the above factors. Generally, non-sampling errors increase with the increase in sample size. Therefore, the size of the sample should be optimized to minimize the sum of sampling and non-sampling errors.
Make sure you also check our other amazing Article on : Friedman Test