Research 101: Measurements of Constructs
0 likes245 views
This document discusses key concepts for measuring constructs in research, including: - Constructs must be precisely defined and some are easier to measure than others like personality. - Conceptualization is defining constructs and their components. Operationalization develops indicators to measure constructs. - Constructs can be one-dimensional or multidimensional. Measurement levels include nominal, ordinal, interval, and ratio. - Common scales for recording assessments include binary, Likert, semantic differential, and Guttman scales. Scaling associates qualitative judgments to quantitative metrics to create measurement scales. Indexes and typologies are composed of multiple constructs through rules or categories.
Research 101: Measurements of Constructs
- 1. Research 101: Measurement of constructs Harold Gamero
- 2. Measurement of constructs • To test theories, it is necessary to be able to measure the constructs of interest in a precise, correct and scientific manner. • Some constructs can be easily measured (e.g. age, gender, firm size). • Others may be considerably more difficult to measure (e.g., creativity, personality, alienation, etc.).
- 3. Conceptualization Definition It is the process by which a construct (imprecise and unclear) and its components are precisely and concretely defined. • For this process it is necessary to understand which elements are included or excluded in the concept. • This process is the most important due to the imprecision, vagueness and ambiguity of social science constructs. • These definitions will not be based on an objective criterion, but on the shared agreement of the mental image that is conceptualized.
- 4. Conceptualization • Sometimes, the constructs used are not real or do not exist independently but are imaginary creations (e.g. prejudice). • According to their complexity, constructs can be: vs. • Constructs that have a single underlying dimension or component. • They can be measured using a single scale or test. • e.g. purchase intentions, self-esteem, age, • Constructs that have two or more underlying dimensions. • Each of the dimensions must be measured separately • e.g. personality, culture, service quality. One-dimensional Multidimensional
- 5. Operationalization Definition Process in which indicators or items are developed to measure a construct. • Due to the high degree of subjectivity, constructs are often measured using multiple indicators. • Multiple indicators allow us to calculate the degree of reliability of a scale. • These indicators operate at the empirical level. • A variable is the combination of indicators that represent a construct. • Each indicator usually has several attributes (response categories), where each attribute represents a value. • These values can be quantitative (numerical) or qualitative (non-numerical).
- 6. Operationalization • There are two types of indicators vs. • They measure the reflection of the presence of a construct. • They measure the construct directly • They measure one-dimensional constructs. Example: Construct: "Religiosity" as degree of religious involvement. Reflective indicator: Number of times the person goes to mass. • The items conform to the construct being measured. • They measure the construct indirectly • They measure multidimensional constructs. Example: Construct: "Religiosity" as a composite of: a) religious belief; b) devotion; and c) presence in rites. Formative indicators: Each of the items of the 3 scales. Reflective Formative
- 7. Measurement levels • According to Smith Stevens (1946), there are 4 generic levels of measurement: Nominal: Also called categorical. Measures mutually exclusive attributes that have no hierarchy. Qualitative. Ordinal: Measures data ordered by ranges but whose values cannot be quantified. Qualitative. Intervals: Measure data ordered by ranges and whose ranges are equidistant from each other. Quantitative. Reason: All of the above qualities and, in addition, with a real zero value (implying the absence of the variable). _ +
- 8. Measurement levels • Statistical properties of measurement scales: Scale Example Central tendency Statistics Nominal Sex, profession, type of industry, religion Mode Chi-square Ordinal Ranking of positions, education level, Likert scales Median Percentiles, nonparametric statistics Interval IQ, final average, degrees Celsius Arithmetic mean, range, standard deviation Correlations, regressions, analysis of variance Reason Age, number of workers, monthly income, degrees Kelvin Geometric mean, harmonic mean Coefficients of co-variation Note: All higher-ranked scales can use the statistics of the lower-ranked scales.
- 9. Scoring scales • Based on the 4 levels of measurement, specific scales have been created for the recording of assessments. • The most used are: a) Binary scales b) Likert scales c) Semantic differential scales d) Guttman scales
- 10. Scoring scales a) Binary scales • Nominal scales, with two response categories. e.g. Political activism scale: Have you ever written a letter to a public official? Yes No Have you ever signed a political petition? Yes No Have you ever donated money to political causes? Yes No Have you ever donated money to a candidate's political campaign? Yes No Have you ever written a political letter to the editor of a newspaper or magazine? Yes No Have you ever persuaded someone to change their voting intention? Yes No
- 11. Scoring scales b) Likert scales • Designed by Rensis Likert. • These are simple statements to which the respondent answers according to his or her degree of agreement or disagreement, on a scale of 5 or 7 points. • Although some researchers tend to eliminate the neutral point, this is a practice not recommended in scientific research. • Likert scales are ordinal in that they are not necessarily equidistant.
- 12. Scoring scales b) Likert scales e.g. Work Self-Esteem Scale Completely disagree Somewhat at odds Neutral Somewhat in agreement Completely agree I feel good about my job O O O O O I get along well with other people in my job O O O O O I am proud of the relationship I have with my supervisor. O O O O O I can tell other people that they are happy to have me at work. O O O O O I can say that my colleagues respect me O O O O O I feel that I make useful contributions to my work O O O O O
- 13. Scoring scales c) Semantic differential scales • Multi-item scale where a single statement has several response alternatives with opposite characteristics. • While in Likert the statements change and the alternatives are the same, in DS the statement remains constant, and the alternatives change. • These scales are the most appropriate for measuring attitudes and feelings.
- 14. Scoring scales c) Semantic differential scales e.g. Scale on attitude towards the public health care system How would you rate the public health care system? Much Something None Something Much Good O O O O O Malo Useful O O O O O useless Insurance O O O O O Insecure Reliable O O O O O Risky
- 15. Scoring scales d) Guttman scales • Designed by Louis Guttman. • It uses a series of items arranged in an order of increasing intensity. • The final score is the weighted sum of the responses (in ascending order).
- 16. Scoring scales d) Guttman scales e.g. Scale of attitude towards immigrants Answer the following questions based on your opinion towards immigrants Would you mind if there were immigrants as citizens of your country? Yes No Would you mind if there were immigrants living in your neighbourhood? Yes No Would you mind if an immigrant is your neighbour? Yes No Would you mind having an immigrant as your friend? Yes No Would you mind if a member of your family marries an immigrant? Yes No
- 17. Scaling Definition Association between qualitative judgments about an unobservable construct and measurable and quantitative metric units. • From this association the "measurement scales" are created. • These scales are empirical structures designed to measure indicators of a given construct. • It is not enough to add punctuation scales to certain statements. • First and foremost is the development of these statements.
- 18. Scaling • Like constructs, measurement scales can be: vs. • They measure constructs with a single scale. • This scale may have several items, but all of them measuring the same construct. • They measure each dimension of the construct separately. • The scores for each dimension can be combined to obtain a total score for the construct. • The scale of each dimension can have multiple items. One-dimensional Multidimensional
- 19. Scaling How to create measurement scales? • Currently there are measurement scales for many of the constructs included in social science research. • Other scales can be modified, translated or adapted, as long as this does not alter their validity. • The best-known methods for creating measurement scales are: a) Thurstone's uniform appearance scaling method b) Likert additive scaling method c) Guttman cumulative scaling method Each of these are explained in detail in the book.
- 20. Indexes Definition They are scores that are composed of multiple constructs (called components) through rules and formulas. • The creation of indices is very similar to the process of measuring constructs: 1) Conceptualization of the index and its components. 2) Operationalization and measurement of each component. 3) Creation of a formula or rule to calculate the index value. e.g. CPI, NSE, Happy Planet Index
- 21. Typologies • While indexes are naturally quantitative, typologies are qualitative. • They are composed of a set of constructs, but the final values are nominal categories. e.g. Types of leadership