ERRORS IN SAMPLING

Answer: SAMPLING AND NON-SAMPLING ERRORS

Errors in statistics are classified in two categories:

1.    Sampling Errors or Random Error        
2.    Non-Sampling Errors or Human Error 

Sampling Errors: 

Sample always gives approximation to the parameter of universe. So, the differences between the actual figure and the estimated figure are always there. These differences are sampling errors. So, the sampling errors always have their origin in sampling. Such errors are not found in census or complete enumeration.

Generally sampling errors are due to the following reasons: 
Improper selection of the sample leads to sampling error. This improper selection may be due to the personal judgment, etc. i.e., non-probability sampling techniques. 

• These errors may be there due to the variability of population and wrong method of estimation. Usually this is in the case of heterogeneous population. 
• Faculty demarcation of statistical units.

Non- Sampling Errors

These kinds of errors are present in both complete and sample enumeration. These errors generally arise when data are not properly observed, approximated and processed. The following factors give rise to the non-sampling errors: 

• Incomplete questionnaire and defective method of interviewing. 
• Errors in compilation and tabulation give rise to non-sampling errors. 

Compilation errors include: 
1. Calculation mistakes. 
2. Personal bias of the investigator. 
3. If the various terms used are not properly defined then it also leads to non-sampling errors.

Measurement of Errors 
 Statistical errors can be measured: Absolutely or Relatively 

Absolute Errors: 
Absolute error is the difference between true value and the estimated value.

Biased and Unbiased Errors:  
 Statistical error can also be divided into the following categories.

1. Biased Errors     
 2. Unbiased Errors

Biased Errors: When the errors are introduced due to the personal bias, these are known as biased errors.  These errors have a tendency to grow in magnitude with the increase in number of observations. 

Unbiased Errors: These are the errors, which do not accumulate with the increase in the size of observations but rather have a tendency to get neutralized. The main purpose of the statistical method is to avoid the biased errors and devise methods in such a way that the errors, if any, are only biased ones. One such devise is random selection over the bias selection.

Total Error : This is the total of sampling error + non-sampling error.  Out of this, the sampling error can be estimated in the case of probability of probability samples, but not in the case of non-probability samples.  Non-samples errors can be controlled through hiring better field workers, qualified data entry persons and good control procedures throughout the project. 

One important outcome of this discussion of errors is that the total error is usually unknown.  But, we may have to live with higher non-sampling error in our attempt to reduce sampling error by increasing the sample size of the study, not to mention the higher cost of a larger sample.  Therefore, it is worthwhile to optimize total error by optimizing the sample size, rather than going blindly for the largest possible sample size.