Chapter-4: Scrutiny of Data

 

 

Scrutinizing data is a vital process in ensuring the accuracy and reliability of research findings. This chapter will discuss methods for checking internal consistency and detecting errors in data collection and recording, with a focus on practical applications and examples relevant to India.

 

 

 

 1. Checking Internal Consistency

 

Definition of Internal Consistency

 Internal consistency refers to the extent to which all parts of the data are consistent with each other. It ensures that the data collected is logical and free from contradictions.

 

Methods to Check Internal Consistency:

1. Reliability Testing: This involves statistical techniques to assess the consistency of the data.

    Cronbach's Alpha: A common measure used to assess the internal consistency of a set of items or questions.

    Example: A survey on customer satisfaction with services in Indian hotels may use Cronbach's Alpha to check the reliability of responses to various servicerelated questions.

 

2. Correlation Analysis: Checking the correlation between different variables to ensure they are logically related.

    Example: In a study on education and income levels in India, one would expect a positive correlation between higher education and higher income. A lack of correlation may indicate inconsistencies in the data.

 

3. CrossVerification: Comparing data from different sources to verify consistency.

    Example: Crosschecking agricultural production data reported by different government agencies in India.

 

Examples:

 In a health survey, if data shows that individuals who report regular exercise also have better health indicators, this consistency supports the validity of the data.

 In a financial audit, if the total assets do not equal the sum of liabilities and owner’s equity, this inconsistency signals potential errors.

 

 

 

 2. Detection of Errors in Data Collection

 

Common Errors in Data Collection:

1. Sampling Errors: These occur when the sample is not representative of the population.

    Example: Conducting a survey on urban internet usage in India but only sampling individuals from metropolitan cities, thereby excluding smaller cities and towns.

 

2. NonSampling Errors: These include all other errors not related to the sampling process.

    Types:

      Measurement Errors: Incorrectly measuring or recording data.

      Example: Misreporting the income of households in a survey due to misunderstanding the question.

      Interviewer Bias: When the interviewer’s actions influence responses.

      Example: An interviewer leading respondents to certain answers in a political opinion survey in India.

      Response Bias: When respondents provide inaccurate answers.

      Example: Individuals underreporting their alcohol consumption in a health survey due to social desirability.

 

Methods to Detect Errors in Data Collection:

1. Pilot Testing: Conducting a pilot survey to identify potential issues in data collection methods.

    Example: Before a national health survey, a pilot test is conducted in a small region of India to identify and rectify any data collection issues.

 

2. Training and Supervision: Ensuring data collectors are welltrained and supervised.

    Example: Training enumerators for the Indian Census on how to accurately record data and handle respondent queries.

 

3. Validation Checks: Implementing validation checks during data entry to catch errors.

    Example: Using software that flags unrealistic entries, such as a respondent’s age being recorded as 150 years.

 

 

 

 3. Detection of Errors in Data Recording

 

Common Errors in Data Recording:

1. Transcription Errors: Mistakes made while transcribing data from one source to another.

    Example: Recording a household’s income as ₹50,000 instead of ₹500,000.

 

2. Data Entry Errors: Errors made during the manual entry of data into a database.

    Example: Typing errors or incorrect coding of responses in a survey database.

 

3. Logical Errors: Entries that do not make sense logically.

    Example: A survey response indicating that a person has completed a PhD but is only 15 years old.

 

Methods to Detect Errors in Data Recording:

1. Double Entry System: Entering the same data twice independently and comparing the entries.

    Example: Entering survey responses into a database by two different data entry operators and then comparing the entries for discrepancies.

 

2. Automated Error Checking: Using software to automatically check for and flag errors.

    Example: Data entry software that flags entries that fall outside expected ranges, such as an income value that is too high or too low.

 

3. Regular Audits: Conducting regular audits of the data to identify and correct errors.

    Example: Periodically reviewing financial records of a company to ensure all transactions are recorded accurately.

 

Examples:

 In a demographic survey, if the total number of males and females does not add up to the total population, this discrepancy signals a recording error.

 In an academic study, if the recorded test scores exceed the maximum possible score, this error must be corrected.

 

 

 

 Illustrations with Examples

 

1. Internal Consistency Example:

    A survey measuring job satisfaction among employees in an Indian IT company might include multiple questions on various aspects of job satisfaction. If the responses to these questions are consistent, it indicates good internal consistency.

 

2. Error Detection in Data Collection Example:

    During a national nutrition survey in India, discrepancies are found in the reported dietary intake due to misunderstanding of portion sizes. A pilot survey helps in refining the questions for better accuracy.

 

3. Error Detection in Data Recording Example:

    In a financial survey, if an individual's monthly expenditure is recorded as higher than their monthly income, it indicates a potential recording error that needs to be investigated.

 

 

 

 References

 

1. "Research Methodology: Methods and Techniques" by C.R. Kothari.

2. "Survey Methods and Practices" by Statistics Canada.

3. Government of India Census Reports.

4. "Data Analysis with SPSS" by Stephen Sweet and Karen GraceMartin.

5. "Statistics for Business and Economics" by Paul Newbold, William L. Carlson, and Betty Thorne.

6. Journal articles and publications from the Indian Council of Social Science Research (ICSSR).