Chapter 3 Flashcards
what does choosing of a measure imply in terms of construct?
implies that you endure view/theory associated with that specific measurement
what is a formal definition of a construct
what your construct corresponds to
what is an operational definition of a construct
how to measure various elements of your formal definition (how to measure your construct)
what is a theoretical model?
refers to how a construct is formally defined and operationalized
what is a measurement model?
examines how responses from individual questions are interrelated
the extent to which the interrelationship among individual questions implied by your theoretical model can be found in actual responses to questions
how is a measurement model different from operationalizing a construct?
operationalizing = involves how a construct is turned into questions/measures
measurement model = involves how numerical answers to questions are scored, and aggregated
what are indicators?
indicate the presence of something (construct)
what does Plato’s cave allegory describe
describes distinction between constructs/measures/indicators
what are the three main points of Plato’s allegory of the cave
we cannot see anything directly (we cannot see things directly, but can make inferences based on indicators)
it is very easy to get things wrong (shadows in allegory are imprecise)
what we observe, even if imprecise, is caused by something real (everything we know is inferred through what we observe)
what is the classical test theory?
1st formal theory of how tests work by Robert Novick (1966) that formalizes Plato’s allegory
states that a persons true score is different from observed score
difference due to error
what equation represents classical test theory?
T = X + C
or
X = T + E
where x = observed score, t = true score, and e = error
what are the difficulties associated with classical test theory?
if we cannot know what the true score is, then it is not possible to calculate the error
if we don’t know how much error there is in the observed score, we cannot calculate the true score
therefore, you have 1 equation with 2 unknowns
how can you approximate true score with classical test theory
if we assume that errors are randomly distributed, then through repeated measure, you can get close to approximating what the true score is
who came up with item response theory
Frederic Lord, Georg Rasch, and Paul Lazorsfield
what is item response theory
explains relationship between an unobservable characteristic/attribute (ex. level of depression) and observational manifestations (ex. responses to test questions)
allows you to estimate how the probability of choosing one of many response options to a question changes with increases in the amount of characteristic/attribute/condition
essentially models the relationship between an individual item and an unobservable characteristic
what are option characteristic curves on item response theory graphs?
allow you to make good guesses about level of condition/ability/characteristic based on response to a single question
how effective are option characteristic curves on multiple choice questions vs. tests that measure different levels of a characteristic/symptom/ability
multiple choice (only one correct answer out of multiple) = not effective/interesting
tests that measure different levels/tests where all options of question should be modelled/are of interest = ideal criteria
what is structural equation modelling?
statistical method used to examine how responses to items are interrelated
allows researchers to convert theory about how individuals questions on a test are interrelated into a set of mathematical equations that describe this
is an example of model testing
what are structure equations in structural equation modelling?
used to mathematically describe the relationship between construct and indicators
what is model testing
model = testing
model testing = testing a theory
theory = model about what world is like/how things work
what is main goal of structural equation modelling?
to turn theoretical statements into a set of mathematical equations with (w actual data) help you understand if theoretical claim is justifiable
how can you tell what structural equation model is the best?
model that best captures observed relationships and is closest to nature of construct/study
what do scientific models represent?
relationships between scores, theoretical models, observed scores, and error
what is unidimensional model
every individual question/indicator can be viewed as measuring sample, single construct
every item/question contributes equally to measuring construct
every error that is introduced through every item score is about the same
relationship among questions that person answers on a test is explained by a single, unobservable entity (construct)
what is non-unidimensional model
states that relationship among questions that a person answers on a test is best explained by various separate and unobservable entities which are both understood at part of the overall single overall unobservable construct
what does modelling data involve?
involves imposing some structure/idea on the data and examining if that model/imposed structure makes sense
what is Ockham’s razor
describes idea that simpler explanations are generally better than more complex ones
if 2 models fit the data equally well, the data is not sufficient to inform which model is best - therefore Ockham’s razor principle should be implemented
what is “fit”
how will a model fits the data
what are the benefits of simplicity?
simpler is easier to understand/explain things
simpler is better when it comes to measuring constructs and predicting outcomes
