Unit 4: Measurement & Scaling
- Measurement
- Definition: Assigning numbers or symbols to characteristics or objects for statistical operations and better communication of results.
- Example: Assigning rupee figures to households with the same income.
- Scaling
- Extension of measurement creating a continuum for measurement placement.
- Example: Satisfaction level from 1 to 11.
- Types of Measurement Scales
- Nominal Scale
- For identification purposes.
- Numbers assigned do not imply superiority or inferiority.
- Examples include religious affiliation.
- Statistical operations: frequency distribution, Chi-square test, contingency coefficient, binomial test.
- Ordinal Scale
- Indicates higher or lower status but not the magnitude of difference.
- Examples: Rankings in a competition, CAT score percentiles.
- Statistical operations: Median, percentile, quartiles, rank order correlation, sign test, plus nominal scale operations.
- Interval Scale
- Differences between scores have meaningful interpretation.
- Example: Rating scales for likelihood, satisfaction, or agreement.
- Statistical operations: Addition, subtraction, mean, standard deviation, correlation coefficient, T-test, Z-test, regression, factor analysis.
- Ratio Scale
- Ratios of measurements have meaningful interpretation.
- Examples: Number of shops, students enrolled, distance traveled.
- All mathematical operations are applicable.
- Classification of Scales
- Single Item Scale: Satisfaction with job, for instance.
- Multiple Item Scale: Satisfaction with aspects of a job.
- Comparative Scales
- Paired Comparison
- Constant Sum
- Rank Order
- Q-Sort and other procedures
- Non-Comparative Scales
- Graphic Rating Scale
- Itemized Rating Scale: Likert, Semantic Differential, Stapel
- Measurement Error
- Occurs when the observed measurement differs from the true state.
- Caused by environmental variations, interviewer bias, ambiguous questions, or coding errors.
- Formula: Observed measurement = True score + Systematic Error + Random Error.
- Criteria of a Good Measurement
- Reliability: Consistency, accuracy, and predictability of a measurement.
- Validity: Free from systematic and random errors.
- Sensitivity: Ability to accurately measure variability in a concept.