Unit 2: Sampling Errors

What is sampling error? How do you minimize it?

Sampling Error:

In sampling, whatever is the method of selection, a sample estimate will inevitably differ from the one that would be obtained from enumerating the complete population with equal care. This difference between the sample estimate and the population value is called the sampling error. The larger the sample the smaller will be the sampling error on the average and greater will be the confidence in the results. Heiman (2002) defined sampling error as "the difference, due to random chance, between a sample statistic and the population parameter it represents".

Sampling error can be of two types, namely, random error and systematic error.

i.                   Random error is a pattern of errors that tend to cancel one another out so that the overall result still accurately reflects the true value. Every sample design will generate a certain amount of random error.

ii.                 Systematic error or Bias, on the other hand, is more serious because the pattern of errors is loaded in one direction or another and therefore do not balance each other out, producing a true distortion.

-------------------------------------------------------------------------------------

Methods to minimize sampling error:

There are various rules by which one can reduce the sampling error. These are:

a.      Using considerably large sample size. As the size increases, the sample gets closer to the actual population, thereby decreasing the potential for deviations from the actual population.

b.      Another potential method of minimizing the sampling error the selection of the sample through probability sampling. Here, every unit of the population has an equal chance of getting selected in the sample, thereby reducing the bias in selection procedure.

c.       Stratification is another method of obtaining greater precision in our sample estimates. In this method, secondary information can be utilized to divide the population into groups such that the elements within the each group are more alike than are the elements in the population as a whole. This ensures greater precision in the estimation of population parameter from the sample statistics.