scale development Flashcards
concrete constructs
connected to unambiguous observable events
usually observed with physical instruments
e.g. weight/height/age
abstract constructs
observed through self-report measures (e.g. scale items)
observe behaviour or psychological measures e.g. brain imaging
e.g. memory/attention/anxiety/identity/happiness
what are constructs called?
latent variables
something we are not directly observing
what are measures called that intend to capture the latent variables?
observed variables
what are 5 ways psychological constructs are measured?
- cognitive tasks
- behavioural observation
- physiological measures
- brain scans
- self-report scales
what type is the focus?
self-report scales
disadvantage of self-report scales (3)
- dependent on pp’s memory e.g. responses can be affected by recency effects (recent event may be more important)
- influenced by self-representation (pp motivated to present themselves in a positive light to others and themselves)
- influenced by pp’s response styles (some pp agree with scale items regardless of their content; have an acquiescent response style)
3 advantages of self-report scales
- information richness (no one has access to as much info about pp than the pp themselves)
- motivation to report (pp usually pleased to talk about themselves and may put in more time and effort)
- ease of administration (cheap, quick, easy to set up, code and analyse)
phases of investigating new construct using scales (Flake et al 2017)
- substantive phase (week 1)
a. construct conceptualisation/literature review
b. generating items - structural phase (week 5)
a. item analysis
b. determining dimensionality
c. reliability - external phase (week 6)
a. convergent/discriminant validity
substantive phase of scale development
make decision about target construction and what the scale will look like
steps:
a. construct conceptualisation/literature review (define boundaries of construct)
b. generating ideas (making measurement decision/review items - quality control)
type of psychological experience: vegetarianism
vegetarian identity: identify as vegetarian
vegetarian behaviour: I generally avoid eating meat
being aware of what experience you are trying to capture when constructing scale items
construct specificity
constructs should be measured at specific level of detail or focus
specificity matching principle states variables at same level of specificity are more strongly related e.g. constructs about life as a whole correlate strongly with other broad constructs compared to narrower ones like workplace related ones
e.g. person may feel like life is meaningful but not find meaning in their career
direct vs indirect items
direct
- pp directly invited to reflect on the constructs of interest and name of construct is used in the item wording
- e.g. “I have high self-esteem”
indirect
- pp indirectly invited to reflect on experiences that are about the construct of interest without naming it
e.g. narcissism scale might ask pp about leadership (“people see me as a natural leader”) but the scale is not about these things
what is an aim for constructs?
good discrimination
- people with same scores should be equivalent in the measured construct
- people with different scores should be different to each other in the measured construct
- the degree of difference between people should be proportional to the difference in scores
examples to understand discrimination
if line A has 4cm and line B has 2cm you would expect line A to be twice as long as B
if pp A scores 4 and pp B scores 2, you would expect A to be twice as afraid of spiders as pp B
number of scale points/response options
more response options = more variability in responses (which is a good thing)
data analysis is mostly about modelling variance between 2 or more variables so if there is no variance in responses, then you can’t model relationships between constructs
BUT too many response options can make it hard to distinguish response options so need to strike a balance
between 4 and 10 options
odd versus even numbered scales
odd numbered scales allow for a mid-point where pp can express uncertainty/not really take a position…but neutral positions may not be good as pp can overuse them to complete scale quickly
even numbered scales force pp to commit to one direct
careful labelling (2 points)
- make response options exhaustive and mutually exclusive (e.g. every once in a while/occasionally/some of the time are all too similar)
- label all your response options (1 - 5 from strongly disagree to strongly agree where all numbers need to be labelled)
brainstorming items (2 points)
- be comprehensive: you should aim to cover all aspects of your target construct with your items
- generate more items than you intend to include in the final scale: some recommend as many as 3 times more than intended number of items
what does redundancy mean?
refers to the presence of items that are highly similar or essentially measure the same underlying concept
redundancy when generating items
can express a similar idea but in a different way
good way:
- I will do almost anything to ensure my child’s success
- no sacrifice is too great if it helps my child success
bad way (too similar)
- a really important thing is my child’s success
- the really important thing is my child’s success
reviewing items: avoid ambiguous terminology
- avoid uncommon words (spendthrift), technical terms, jargon, slang, acronyms (e.g. UI
- aim for wording that all respondents will interpret in the same way (e.g. sanction can mean either to authorise something or to punish someone)
reviewing items: write clear items
- avoid double negatives (e.g. are you against the ban of smoking?)
- avoid double-barrelled items (split video games and board games)
- avoid negative items (original item is “I like going out on Friday” then change to “I like staying in on Friday” rather than the negative “I don’t like going out on Friday)
