Week 2 - item development and test construction Flashcards
Terminology
- Scale (or measure): A series of items that assess a particular construct.
- Item: Each individual statement/question within a scale/measure
- Factor: A group of items that reflect an underlying construct, or an aspect of that construct.
What is the first part of measuring a psychological construct?
Defining the test
•What is the construct to be measured?
•Who is the new test/measure/scale intended for?
•What is the intended aim or use of the new test/measure/scale?
Example: After an extensive search of the pizza-related literature, there seem to be two important aspects•Taste •Smell•Items will need to be developed to separately assess how much people like the taste and the smell of pizza
What are the scaling methods?
What do you need to measure?
•Stevens’ typology:
- Nominal
- Ordinal
- Interval
- Ratio
- Measures of constructs tend to be an interval
- Ordinal and nominal scales are rare (but often part of surveys to assess factual or demographic information)
- Ratio is rare (i.e., requires a meaningful zero point. It is possible to have zero weight, but is it possible to have zero intelligence?)
To briefly explain the forms of measurement:
•Nominal: Generally a category where there’s no clear ranking (e.g., gender identity, nationality).•
Ordinal: Categorical data with a clear ranking system. E.g., if people have to respond to the question ‘How is your health?’ be selecting one of the following responses: Poor, Average, Excellent
•Interval: Any data where the difference between any two values is equal. E.g., temperature –the difference between 10 and 11 degrees Celsius is the same as the difference between 100 and 101 degrees Celsius•Ratio: The same as interval data but with a meaningful, absolute zero point. E.g., length measured in centimeters.
What are some of the formats?
Item formats
- Nominal:•Multiple choice format. E.g., asking about gender:•Male•Female•Nonbinary•Prefer not to say•Other (with an open text box for a respondent to enter their own response).
- Ordinal (e.g., asking about educational experience):0 = Did not complete high school 1 = Completed high school 2 = Incomplete undergraduate studies 3 = Complete undergraduate studies
Rating scales
•Guttman scale
- An ordered set of items such that agreement with an item means agreement with all preceding items
- A scaling technique, but sometimes such as an item format•E.g., a measure assessing depressive affect:
a) I occasionally feel sad
b) I often feel sad
c) I feel sad most of the time
d) I always feel sad
Rating scales
Respondents assign a number to represent the magnitude of an attribute.
•Likert scale is the most common•Technically has 5 possible responses, but it’s common to find ‘Likert-type’ measures that use a different number of response options.•E.g., Satisfaction With Life Scale:
Rating formats
Dichotomous
•True/False•Agree/Disagree•Yes/No
•Can be an easy way of identifying factual beliefs, or opinions/attitudes
•Two options can be limiting and can make it easier for socially desirable responses to occur.