Final Exam Flashcards
What is a variable?
A characteristic on which people differ from one another
Does a variable need to be measured?
Yes
What is an observed variable?
Variable that can be directly assessed (age, sex, height)
What is a latent variable?
Cannot be directly measured
What is a psychological construct?
Is an abstract entity that was created to reflect a set of behaviors that tend to co-occur with one another
How is a psychological construct assessed?
By getting a representative sample of these behaviors, this sample should be both limitative and inclusive
What is related to the degree to which the sample of behaviors is representative of the construct?
The quality of the measurement (the diagnostic value) and the ability to generalize the conclusion (predictive value)
What is the test designed for?
To solicit the sample of behaviors
What is the most typical assumption made in Psychology?
The presence of the latent construct predicts the emergence of the observed behaviors
What method do we use to estimate latent variables?
Factor analysis
What is confirmatory factor analysis?
- When the researchers determine in advance which specific items go with which specific factor
- Check if representation matches the data
What is exploratory factor analysis?
- Feed all items in analysis to see which items have a higher association with which factor
- Nature of the factors comes from all the items
What is the difference between factor loading and cross loading?
Factor loading is the strength of the association between item and factor whereas cross loading is a weaker loading of the association
What is the formula for factor analysis?
X = Tau + Lambda *Xi + Delta
Observed score = intercept + regression slope * score on predictor + residual
What is the goal of a factor analysis?
To analyze a set of continuous observed variables in order to:
- See if they form a relatively independent and meaningful subset
- Understand the underlying structure/organization of the set of variables
- Synthesize a larger set of variables
Why are factor analyses a critical component of psychometric validation studies?
To see whether the various items on the questionnaire actually assess the underlying construct (factor)
What two types of ways can a factor analysis be conducted?
Using ordinal (specialized application) or nominal (conducted analyses)
Different characteristics leading to factor analysis
- The indicators have two causes: the factor and the random measurement error
- The residuals/uniqueness describe what is unique per indicator and is not shared among other indicators
- What is shared between the indicators is absorbed in the factor
- The latent factor is corrected for measurement error and is perfectly reliable
- Conditional independence: factor analysis “assumes” that all the covariance is absorbed by the factor (no residual correlations among the uniqueness)
Structural equation modeling?
Provides a way to estimate relations among constructs corrected for random measurement errors
Factor analysis
- Focuses on covariance: what is shared among the indicators
- Reflective model: the indicators are seen as providing a reflection of the latent variable
- The indicators have 2 causes = factor and the uniqueness
Principal component analysis
- Looks at the complete variance-covariance, thus what the indicators share and what is unique among them
- Formative model: the indicators form the latent factor
- Provides a summary index of unrelated indicators
- Assumes that you are interested in all that is the indicators
What is a bias?
The presence of a systematic difference in the estimation of scores on a given construct (=validity) for a given group. A bias is present when the margin of error differs systematically from one group to another
Can you get a test bias if you find the mean difference?
No, when one group scores on the average higher than another groups it is not a bias, groups may truly differ from one another on the construct being assessed
What are the different types of bias?
Content validity bias, construct validity bias, predictive validity bias, slope bias and intercept bias
What is content validity bias?
- When the item or subscale is systematically more difficult for one group compared to another, holding constant the individual skill level
How to avoid content validity bias?
- Create a subsample with comparable scores on test within each cultural group
- Within subsample assess item difficulty level to see if it differs across cultures
What is construct Validity bias?
When the test assess different things in different groups or identical thing with different degrees of precision (factor analysis)
What is predictive validity bias?
When the quality/precision of the prediction differs across groups
What is slope bias?
The most severe form of bias, when the coefficient of validity is not the same across the groups
What is intercept bias?
When the intercept differs across groups. i.e; equivalent score on the test systematically predicts higher level of performance in one group relative to the other, this type of bias is easy to fix when identified
Advantages of the interview
- Allow more depth and flexibility (less rigid more humane)
- helps to build rapport with client in clinical context
- allows to collect a lot of information with minimum level of preparation
- substantial cost-benefit ratio when no appropriate questionnaire exists or few persons are assessed
- a psychological assessment must include an interview
Disadvantage of the interview
- Hard to control for bias due to assessor
- Person being interviewed will be more susceptible to social desirability biases and influenced by assessor
- More flexibility means less standardization
- More time is required to collect a similar amount of information
- With many participants the costs associated with interviews are important
Types of interviews (from standardized to more flexible)
Group test, computerized tests, individual tests, structured interviews (explicit question, structures, marking guidelines, stem and follow up questions), semi-structured interviews, unstructured interviews (panel interviews limit biases associated with interviewer)
What 2 essential objectives do clinical interviews have?
1) Conduct a precise and complete assessment of the clients problem, consultation motives, expectation, life functioning and factors that help maintain the problems
2) Building rapport with client to develop strong therapeutic relationship