Topic 10: Measurement and Scaling Flashcards
What is measurement?
Process of describing some property of a phenomenon by assigning numbers in a reliable way
How do you determine what needs to be measured to address a research question?
The problem definition process should determine which constructs/variables should be measured (Decision statement→Research questions→Hypotheses)
How do you measure concepts?
You operationalize the variables (identify scales that correspond to variance in the concepts.
Variables are concepts who’s values can change and be measured
What are constructs?
Concepts that can be measured in multiple ways with multiple variables
What are the different levels of scale measurement?
- Nominal scale- assigns different arbitrary values to items purely to differentiate between them (Most elementary level of measurement, e.g: numbers of different race horses)
- Ordinal scale- assigns values to rank items in ascending or descending order (e.g: Top 5 horses in order of the place they finished)
- Interval scale- Have both nominal and ordinal quantities while also showing the amount of variance between different items (e.g: ranking the top 5 horses and showing what their race times where)
- Ratio scale- All the properties of interval scales and has a meaningful absolute zero value. (e.g: prices have an absolute zero, whereas temperature does not, 0 degrees does not equal the absence of heat)
What are the 3 criteria for good measurement?
Reliability- how consistently does a measure provide the same results (assess reliability using the coefficient alpha method or the test-retest method)
Validity- a measures accuracy; extent to which a score truthfully represents a concept
Sensitivity- a measurement’s ability to accurately measure even small variance in results
With regards to the requirements for good measurement
How is validity established?
- Construct validity (reliably measures and truthfully represents concept)
- Construct validity consists of:
- Face validity: results look realistic and appear at surface value to logically represent what was being measured
- Content validity: does the measure adequately cover the domain of interest (not more or less than what is being measured)
- Convergent validity: concepts that should correlate do in fact correlate
- Discriminant validity: how unique is the measure? A scale shouldn’t correlate too strongly with a measure of a different concept. (If they do, then one must ask whether they are actually different concepts or the same thing)
What is an attitude?
How someone responds to aspects of the world (influenced by cognotive, affective (emotional/feeling) and behavioural components
What are the techniques for measuring attitudes?
- Direct verbal statements used to measure behavioural intent (ranking, rating, sorting, choice)
- Qualitative techniques
- affective(emotional) components of attitudes can be measured using physiological measures (heart rate, pupil dilation)
- attitude rating scales
What are the different types of rating scales and their characteristics?
- Category scale- indicate response from list of categories (easy to respond to, but lack of specific distinctions between options)
- Likert scale- scale of agreement/disagreement (easy to make, difficult to judge individual score meanings)
- Semantic differential & numerical scales- choose points on a scale inbetween 2 bipolar adjectives (easy to make, but bipolar adjectives must be found)
- Stapel scale- points on a scale with single adjective in the centre (don’t need bipolar adjectives, end points are numerical)
- Constant sum scale- divide total number among response alternatives (approximates interval measure, but complex to complete)
- Graphic scale- choose point on a continuum (visual impact, unlimited scale points, but no fixed answers)
What are latent constructs and how would you represent them?
- Use of a likert scale that has multiple scale items to be rated
- After reverse coding if neccessary, all the respondent’s reactions are summed together to form a composite scale to represent all variables of the latent construct
Questions to ask when selecting a measurement scale?
- Is ranking, sorting, rating or choice the best technique?
- Should a monadic(1 variable) or comparative scale(2 variables) be used?
- How many categories/response positions are needed to accurately measure an attitude?
- Balanced or unbalanced rating scale?
- Even or odd number of options?
- Should there be predetermined options to choose from?
- Single measure or index measure?