W7: Longitudinal designs (structural equation modelling) Flashcards
Possible MCQ Question:
What is one statistical advanced method that we can use to test things longitudinally?
Structural equation modelling
Longitudinal designs are … (2)
Used to study changes across the life span by observing same participants at different points in time
It assess same variables at each timepoint
The length of the longitudinal design study can vary
Based on design or funds (longitudinal designs are expensive)
Longitudinal designs are often
survey-based (e.g., Likert scale questions)
Longitudinal designs prone to
many threats of internal validity (e.g., maturation)
Longitudinal designs require
Substantial resources and time to complete
Longitudinal design (issues)
What is time lag?
Distance between timepoints
Longitudinal design (issues)
Too much time lage can
causes issues
Longitudinal design (issues)
Issue with time lag - frequency
How frequent should we measure?
Longitudinal designs (issues) - Attrition (3)
- Participants dropout over time
- Loss of interest
- Losing contact with researcher (e.g pps changing phone number, home address)
Longitudinal design (issues) - Attrition - how do we address this?
Money for the participants to engage in the study/Amazon vouchers
Longitudinal designs (issues) - threats to the internal validity of long study due to - (3)
- Maturation effects (naturally occurring process)
- Historical effects (events that are outside of our control) - e.g., COVID-19
- Practice effect (improvement on test due to repeated experience)
How do we test longitudinal studies? (repeated-measure design)
Repeated measures t-test and ANOVA
More complex longitudinal design requires more robust statistics (than repeated measure t-test and ANOVA)
such as… (3)
- Mediator/mechanism (3rd var)
- Understand complex pathways to poor health
- Enter… Structural equation modelling!
Structural equation modelling - Most longitudinal studies utilise survey designs (7)
Scales which include several items
Rated using Likert scales
- 1 (Strongly disagree)
- 2 (Disagree)
- 3 (Neither agree or disagree)
- 4 (Agree)
- 5 (Strongly agree)
Structural equation modelling - what are construct?s
Hypothetical variable we are trying to capture
Structural equation modelling
Construct limits (2)
Almost impossible to actually measure directly
Assess variables which estimate the construct
Structural equation modelling -=
We can capture the constructs through… (4)
Self-report
Behavioral
Physiological
Etc
Structural equation modelling - construct with its operational definitions
- E.g., may measure aggression (construct) through behavioural observations (e.,g operational definition)
Structural equation model tests
“fit” of each item on construct
Structural equation model estimates a
“weight” for each path - how much each item predicts the construct
Structual equation model interprets the weight as a
standardized regression coefficient (beta) or correlation
Structural equation modelling high and low weights (3)
Higher weights = good fit!
Lower weights = poor fit!
The better the fit, the more predictive that item is of the construct!
Structural equation modeling
- 3 best predictors of anxiety is 1, 3 and 5 (green)
- Red is maybe not best predictor of generalised anxiety - tells us maybe not measuring anxiety
Structural equation modelling
Using constructs to predict .. constructs (2)
Once we have items that nicely fit onto a construct, we can then use it to predict other constructs with good fitting items
Can create pathways based on evidence
Example of pathways (3) with different constructs
Structural equation modelling
From theory to testing… (2)
Structural equation modelling often used to test theories
Provides evidence on whether theories are accurate or need revision (whether pathway someone theorised exists or not)
Racism and health
Gibbons et al. (2014) - PROCEDURE (3)
Longitudinal study with 680 female Black Americans
Assessed 4 times over several years
Used structural equational modelling!
Racism and health
Gibbons (2014) - Measures (5)
- Perceived discrimination
- Internalizing factors (depression, anxiety, distress)
- Externalizing factors (hostility, anger)
- Physical health
- Substance use (alcohol)
Gibbon et al. (2014)
Structural equation modelling (3)
Use of evidence and theory to form a structural equation model
Evidence shows that..
Internalizing (distress) predictive of poor physical health
Externalizing (hostility) predictive of greater alcohol use
Gibbon et al. (2014)
If… (2)
If racism leads to distress (e.g., depression, anxiety)… - makes them have poor physical health
If racism leads to hostility and anger…leads to greater alcohol use
Gibbons et al. (2014) findings (4)
Racism leads to distress which predicts poorer physical health
However… while racism promotes more hostility, this does not necessarily lead to greater rates of substance use!
- Contrary to the proposed model
- Not good enough fit!
Gibbons et al. (2014) - Limitations
Men are on average more hostile and also tend to have more problematic alcohol use than women
Could explain why the pathway of Time point 3 of hostility between Time 4 problematic drinking use
