Study guide: Preparing and Processing Primary Data

📋 Course Outline

1. Purpose and Sequence of Data Preparation
2. Validating Survey Data
3. Editing Questionnaires and Responses
4. Coding Responses and Building Codebooks
5. Capturing and Sourcing Data
6. Web Questionnaires and Database Data
7. Importing Data into Analysis Packages
8. Verifying and Cleaning Data
9. Labeling and Storing Prepared Data

📖 1. Purpose and Sequence of Data Preparation

★ Must-know

• Primary data preparation converts raw questionnaire responses into useful information by validating, editing, coding, capturing, importing, verifying and cleaning, labeling, and storing the data before analysis.

Further detail

• For a large project, preparation and processing can proceed as fieldwork data becomes available rather than waiting for fieldwork to finish.

💡 Memory Hook

Validate, edit, code, capture, import, clean, label, then store.

📖 2. Validating Survey Data

🔑 Key Concepts & Definitions

• Data validation: The process of ensuring that gathered data is valid and accurate and deciding whether each questionnaire should be included in the analysis or discarded.

📝 Essential Points

📌 A questionnaire is included in the analysis only if it results from an interview conducted correctly and consistently.

• Validation guidelines include:
  • Telephone between 10 and 20% of listed respondents to confirm they were interviewed.
  • Check the questionnaire and interviewing instructions against sample requirements.
  • Review each interviewer’s call sheet for correct sampling procedures.

📖 3. Editing Questionnaires and Responses

🔑 Key Concepts & Definitions

• Editing: The critical checking of questionnaires and observation forms for correctness and completeness, with adjustments made where necessary and worthless forms discarded.

★ Must-know

📌 Field editing is preliminary editing by a field organiser or supervisor, whereas central office editing is a more thorough scrutiny and adjustment, usually at the research body’s head office and preferably by one expert.

• Editors assess:
  • Interviewer cheating.
  • Compliance with sampling requirements.
  • Relevance of answers.
  • Completeness of questions and sections.
  • Comprehensiveness and unambiguity of answers.
  • Comprehensibility of answers.
  • Legibility and clarity of handwriting.
  • Inconsistencies.

Further detail

• The researcher can:
  • Review questionnaire and interviewer quality.
  • Return to the interviewer or respondent for better responses.
  • Discard unsatisfactory parts of the questionnaire.
  • Discard the entire questionnaire.

💡 Memory Hook

Field editing is preliminary; central office editing is more thorough.

📖 4. Coding Responses and Building Codebooks

🔑 Key Concepts & Definitions

• Coding: The categorisation of responses into classes and allocation of code numbers to those classes so raw data can be represented in computer-readable form and tabulated.
• Codebook: A manual giving general coding instructions and describing each variable, its code name, its location in the computer record, and how it should be read; no standard compilation procedure is specified.

📝 Essential Points

• Coding proceeds by specifying a separate category for every possible response, assigning code numbers to the categories, and compiling a codebook describing the variables and their coding.

📌 Pre-coding assigns predetermined response categories when the questionnaire is designed, whereas post-coding codes data after collection is complete.

💡 Memory Hook

Categories → codes → codebook.

📖 5. Capturing and Sourcing Data

🔑 Key Concepts & Definitions

• Data capture: Entering questionnaire codes into an electronic data file that a computer can read for analysis.

★ Must-know

• Data may come from:
  • Paper-based survey questionnaires.
  • Web-based questionnaires.
  • Experimental designs.
  • Databases.
  • Optical scanning.

Further detail

• Data-capture formats and tools include:
  • Word-processing text files such as MS Word.
  • Spreadsheets such as MS Excel.
  • Statistical packages such as SPSS or SAS.

📖 6. Web Questionnaires and Database Data

★ Must-know

📌 Web questionnaire studies should ensure respondents are unique, account for internet-access bias, consider combining web and paper questionnaires, and set a submission cut-off date.

• Web questionnaires can be administered effectively and inexpensively because no intermediary interviewer is involved; responses populate a database and can be downloaded after the survey in an electronic format such as MS Excel.

Further detail

• A web questionnaire is designed online, its link is emailed to respondents, respondents complete and submit it to a database, and the completed survey data can then be downloaded electronically.

• When using database data:

  • Carefully choose which data to extract.
  • Clean and validate the data.
  • Remove duplicates.
  • Explain how the database was designed.

📖 7. Importing Data into Analysis Packages

★ Must-know

• Before analysis, electronic data from sources such as spreadsheets, web questionnaires, or scanning must be imported into an analysis package.

Further detail

• Named statistical analysis packages include:
  • SAS.
  • SAS/JMP.
  • SPSS.
  • S-Plus.
  • STATA.
  • Statistica.
  • Minitab.

📖 8. Verifying and Cleaning Data

🔑 Key Concepts & Definitions

• Data cleaning: Checking and correcting data-entry errors by testing internal consistency, including possible and impossible codes.

★ Must-know

• Data-entry errors can include:

  • Impossible values, such as 9 on a scale coded from 1 to 7.
  • Transformed values, such as entering age 52 as 25.
  • Duplicate foreign data.
  • Missing data.
  • Records out of sequence.

• To check for errors, calculate frequencies for nominal and ordinal variables, means and maximum and minimum values for continuous variables, and frequency cross-tabulations for inconsistent or contradictory responses.

📌 A missing value is an incomplete field and is not the same as zero; statistical packages exclude missing values from calculations, commonly representing missing numeric values with a dot and missing categorical values with a blank space.

Further detail

📌 If a value appears incorrect, researchers should consult the original questionnaire when uncertain rather than simply removing the data, and may use the statistical package’s recoding facility to correct errors.

• Cleaning tasks include checking correct codes, inconsistent or contradictory responses, extreme answers, and missing data.

💡 Memory Hook

Missing data is not zero; a missing value is excluded from calculations.

📖 9. Labeling and Storing Prepared Data

🔑 Key Concepts & Definitions

• Variable label: A readable name assigned to a questionnaire question represented as a variable, such as labeling Q1 as Province.
• Value label: A readable description assigned to a variable’s coded category, such as 1 = Western Province and 2 = Northern Cape.

★ Must-know

📌 Raw data preparation is complete once data has been entered, verified, cleaned, and labeled; the prepared data can then be stored in the statistical package format in computer memory or on storage media until analysis.

Further detail

📌 Statistical calculations use numeric codes rather than value labels, while analysis output displays variable and value labels.