PRC Data Quality & Validation 2 — Questions and Answers
Question 1: Which of the following best describes 'data cleaning' in the research process?
- Removing all open-ended responses before analysis
- Identifying and correcting errors, inconsistencies, and outliers in a dataset prior to analysis (Correct answer)
- Deleting records from underrepresented demographic groups
- Applying sampling weights to correct for design effects
Correct answer: Identifying and correcting errors, inconsistencies, and outliers in a dataset prior to analysis
Data cleaning encompasses the systematic process of detecting and correcting inaccurate, incomplete, or inconsistent records to improve data quality before analysis.
Question 2: A researcher discovers that 5% of survey records in a dataset are exact duplicates. The most appropriate corrective action is to:
- Average the responses from duplicate records before analysis
- Retain all duplicate records to maximize sample size
- Remove duplicate records, keeping only one instance per unique respondent (Correct answer)
- Flag duplicates with a weight of 0.5 each
Correct answer: Remove duplicate records, keeping only one instance per unique respondent
Exact duplicate records represent the same respondent counted multiple times, which inflates sample size and biases results; removing all but one instance corrects this.
Question 3: Which of the following is a common method for handling missing data that involves predicting missing values based on other variables in the dataset?
- Listwise deletion
- Pairwise deletion
- Multiple imputation (Correct answer)
- Mean substitution
Correct answer: Multiple imputation
Multiple imputation creates several plausible replacements for missing values based on observed data patterns, preserving statistical power and reducing bias better than deletion methods.
Question 4: In data validation, a 'range check' ensures that:
- All respondents fall within the target demographic
- Numeric values fall within predefined acceptable boundaries (Correct answer)
- The survey was completed within the specified fieldwork dates
- Response distributions match expected population parameters
Correct answer: Numeric values fall within predefined acceptable boundaries
A range check verifies that all numeric entries fall within logically acceptable minimum and maximum boundaries (e.g., a 5-point scale should only contain values 1–5).
Question 5: The practice of verifying that a percentage distribution of responses closely mirrors known population benchmarks is called:
- Internal consistency checking
- Demographic calibration validation
- Benchmarking or external validation (Correct answer)
- Respondent profiling
Correct answer: Benchmarking or external validation
Benchmarking or external validation compares survey marginals against known population statistics (e.g., census data) to assess whether the sample reflects the target population.
Question 6: Which of the following quality indicators would MOST directly reveal that an online panel has a problem with professional survey takers ('professional respondents')?
- High incidence rate for the screening criteria
- Respondents claiming membership in implausibly many product categories or surveys (Correct answer)
- A break-off rate above 20%
- Low item non-response on attitudinal scales
Correct answer: Respondents claiming membership in implausibly many product categories or surveys
Professional respondents often claim to own or use many products and services or belong to many survey panels in order to qualify for numerous surveys, making implausible membership claims a key red flag.
Question 7: What is the primary benefit of conducting a pilot test or soft launch of a survey before full-scale fielding?
- It permanently locks in the final questionnaire so no further changes can be made
- It allows researchers to identify data quality issues, technical errors, and question problems before full deployment (Correct answer)
- It generates a representative sample that can be merged with the main study data
- It replaces the need for respondent validation checks during the main study
Correct answer: It allows researchers to identify data quality issues, technical errors, and question problems before full deployment
A pilot test or soft launch uncovers problems such as confusing wording, skip logic errors, and poor response distributions before they affect the full sample, allowing corrections to improve data quality.
PRC Data Quality & Validation 2
Which of the following best describes 'data cleaning' in the research process?