What is Sampling Errors ?
Sampling survey refers to the process of analyzing small percentage of the total population or data available, basis which, conclusions may be drawn for the entire population. Since the process of sampling does not account for all units of a population, it creates an environment of inaccuracies, and the process of data collection and its study can also be erroneous.
Types of Errors
Errors can broadly he classified into the following two types :
- Sampling errors
- Non-sampling errors
What is Sampling Errors ?
Errors that arise due to variations in collected samples or due to differences between the collected samples and the population at large are referred to as 'Sampling errors'. These errors arise at the first step of the sampling survey procedure, that is, collection. of samples. Sampling errors arise mainly because the surveyors make use of a small population of the larger database, to draw conclusions.
Types of Sampling Errors
Sampling errors can further be classified into the following two types :
1) Biased Sampling Error :
Biased sampling error, as the name suggests, may arise due to certain preferences at the time of collection of samples.
2) Unbiased Sampling Error :
In some cases, collection of samples is to be done keeping a set of conditions in mind. However, it is also required that such conditions do not bring about a bias in the process of sample collection, failing which, an error arising out of such bias is termed as 'Unbiased sampling error.
Causes for Sampling Errors
Sampling errors may arise due to the following reasons :
1) Faulty Selection, Process :
In some cases, a certain kind of samples may be selected that will help reach a desired result. This bias in selection of a sample can cause sampling errors.
2) Substitution :
There may be cases where a collected sample may not be able to enumerate the results that are desired. In such cases, the surveyors may substitute a particular sample member for another similar member that may lead to inaccurate results.
3) Faulty Demarcation of Sampling Units :
A faulty method of separation of sampling units may also lead to such an error.
What is Non-Sampling Errors ?
Non-sampling errors are those that may arise after the process of sampling is complete. Such errors arise at the time of study or analysis of sample data and can occur at any time through the procedure. Such errors occur with both the methods i.e., census as well as sample method of research.
Types of Non-Sampling Errors
Following are the various non-sampling errors :
1) Frame Error :
All the elements of the target population, which can be selected to form a sample are called sampling frame. An error arising due to an incomplete or inaccurate sample frame may be defined as a frame error. For example, if a survey is to be conducted to determine the average household income in a state, it may not be appropriate to collect information about the accumulated salaries of such households alone, as there may be other methods for a household to earn income that is separate from their regular source,
2) Non-Response Error :
It is almost impossible to obtain data from each and every respondent covered in the same. There are always some respondents who refuse to give any information. Thus, non-response error occurs when respondents refuse to cooperate with the interviewer by Sampling errors may arise due to the following not answering his questions. This error also occurs when respondents are away from home when the interviewer calls on them. In case of mail survey particularly, the extent of non-response is usually high.
In a data collection process, it is not sure that absolute response can be collected, through every respondent. Non-response error may occur when respondents are either unavailable, or do not co operate with surveyors at the time of data collection. Such errors usually occur in cases where persons are requested to respond via mall or they are not at their homes, Nan-response error is high in mail surveys.
3) Measurement Error :
At the time of data collection, a respondent may not necessarily give a true picture, due to innumerable reasons. This leads to an inaccurate result as the information that is collected is not based on facts.
For example, if data is being collected to determine the average household income of a state, the respondent may not be willing to give his/her actual income or for data of number of youngsters who smoke in a city, majority of respondents may negatively respond. This will, in turn, lead to an inaccurate result.
4) Data Processing Error :
Data processing refers to the process of systematic categorization of data to make the process of analysis easier and more accurate. However, errors may occur at the time of categorizing data, such as, drawing up of tables, coding responses, etc.
5) Data Analysis Error :
Data analysis errors may be defined as those errors that arise due to the application of incorrect statistical techniques or formulas that give the wrong result. These errors may be simple as well as complex.
Causes for Non-Sampling Errors
Some of the important causes of non-sampling errors have been defined below :
1) Faulty Planning and Definitions :
Non sampling errors arise due to improper definition of samples, lack of trained surveyors, improper definition of target population, etc.
2) Response Errors :
Response errors may be defined as errors that arise due to variations in responses of participants from actual facts.
3) Non-Response Errors :
Following may be the reasons for non-response errors :
- The respondent is unavailable even after repeated efforts to get in touch with the same.
- The respondent is unable to respond to all questions or provide information pertaining to the same.
- The respondent is not willing to answer some or all of the questions.
4) Errors in Coverage :
Lack of ability to cover all sampling units gives rise to such errors.
5) Compiling Errors :
Compiling errors are those that arise due to erroneous editing and coding of data by the researcher.
Methods to Reduce the Errors
The following techniques can prove helpful in reducing the errors stated above, both sampling and non-sampling :
Methods to Reduce the Sampling Errors
For sampling errors, the following techniques may be applied :
1) Increasing the Sample Size :
One of the easiest methods of reducing sampling errors is to increase the size of the population. Sampling error may be zero in case of sample size n being equal to population size N. Square root formula reduces the error percentage by half when the sample size is increased to four times its original.
For example, if in a sample size of 1,000 units the error percentage is 10%, an increase of this sample size to 4,000 units will lead to a reduction in the error percentage to 5%.
However, if the collected samples are unbiased, then a decrease in the sample size will also lead to a decrease in the error percentage, as this decrease will be inversely proportional to the square root of the sample size; sometimes square root of the sample size is inversely proportional to the extent of decrease.
2) Stratification :
Stratification may be defined as the process of dividing the samples of a similar kind into a particular group or strata. From each of these groups or stratum, a sample will then be collected randomly. This technique should be applied to cases where sampling units are varied in nature and the collection of these sampling units on the basis of random sampling will fail to represent the population as a whole. Thus, stratification will help to reduce errors as all groups of a population will be available for sampling. This technique is often also described as stratified-random sampling. The size of the sample collected from each stratum is generally proportionate to the size of its stratum.
This is also known as fixed sampling fraction and is often used in sample surveys. However, one drawback of this technique is that it requires prior availability of data regarding population units, failing which it will not be possible to use the same.
Methods to Reduce the Non-sampling Errors
For non-sampling errors, the following techniques may be applied :
- Responsible collection of samples at appropriate times,
- Use of an accurate sampling frame,
- Appropriate plan for following-up on non respondents,
- Designing a comprehensive questionnaire,
- Thorough training of surveyors and data processing personnel,
- In-depth knowledge of factors that affect the research problem.
Recommended Terms :