Please enable JavaScript.

Coggle requires JavaScript to display documents.

Research Methods M2 (Survey Research (Developing a Good Questionnaire…

- - - - Multiple Choice
        
        Single Response: Used for priority; asking for one response out of several
        
        Multiple Response: Ask to select all answers that apply, or rank them in some way (more freedom)
      - Force Choice: Participants pick a single response that reflects their priority/belief/attitude out of a set of options
      - Likert Scale: Participants rate the degree to which they agree or disagree with a statement
      - Scenarios: Can be any of the 3 above where participants read a paragraph about a situation and answer questions about how they would feel/react
      - Semantic Differential: Participants describe their feelings on a scale
        
        Not every number is labelled
      - Visual Analogue: Uses a straight line/slider with each end labelled that can be marked anywhere along it
- - - - Diener (1985) measured this subjective happiness with a "satisfaction with life" survey, SWLS. He operationalized happiness with a measure of life satisfaction
    - - Diener (1985) had a measure for happiness, he believed people were aware of how happy they are so his conceptual definition for happiness was the subjective well-being of participants; how happy they think they are
  - - - Observational measures: operationalize variables by recording either observable behaviours or physical traces
        
        Physical traces: What people leave behind (used to investigate behaviours); like shortcuts across a grass field
        
        One could operationalize happiness by observing smiling frequency
      - Physiological measure: Record one's biological data
        
        Like observing cortisol level or heart rate when operationalizing stress
      - Self report measure: Have people answer questions about themselves
        
        Like surveys, best for situations where you can trust a person's self report
    - - Quantitative
        
        Ordinal: Shows relative positions but not the size of gaps in between levels; can't do inferential stats with this
        
        Interval: Equal gaps between numbers but no meaningful 0, like temperature, IQ, or a likert scale
        
        Ratio: Equal gaps between numbers with a meaningful 0, like reaction time or weight
      - Categorical: Represents qualitative differences
        
        Nominal: Membership is meant to be mutually exclusive. The weakest scale used for categorical data
- - - - Cross-Sectional: The correlation between different variables at the same time
      - Autocorrelations: The correlation between the same variable at different times
      - Cross-Lagged: The correlation between different variables at different times
    - - Partial Correlations: Reduces the variance due to some 3rd variable, z, to see if it mediates the relationship between x and y
        
        Mediating Variable: A variable that explains the relationship between two other variables
        
        If a relationship is significant: "X is significantly (positively/negatively) correlated with Y independent of Z"
      - Multiple Regression: A statistical technique that observes the effects of 3rd variables by observing all correlations while controlling for all other variables
        
        Choose your DV. All other variables are considered predictor variables
        
        Beta tells us the direction of the correlation
        Significance is also provided
        
        "X is significantly (positively/negatively) correlated with Y independent of Z1, Z2, Z3..."
  - - - easier to make causal claims when you do an experiment based on correlational findings: we can establish temporal precedence and control for everything with random assignment
      - ex. with recess study, you can assign classes to have more or less recess, obviously hiding the hypothesis from the teachers (good way to make a compelling argument to the gov)
  - - - Correlation =/= Causation
        
        Directionality Problem: If temporal precedence was not established, we are unaware if x caused y or the reverse
        
        Can be addressed logically; if y could not possibly cause x, then there is no directionality problem (ex. weather vs attitude: attitude could not possibly cause weather)
        
        Third Variable Problem: Some other variable z may be responsible for the correlation between x and y
        
        Third vs. Confound: Third variables are variables we failed to measure that could explain the relationship, confounding variables are alternate explanations for causal research
      - Coefficient of Determination: The proportion of variation in Y due to X
- - - - Content: Has the "right" items and assesses all aspects of the construct
      - Face: Does it look like it measures what it's supposed to?
    - - Criterion: The measure is able to predict how someone will do on some concrete criterion (is the measure correlated to what it's supposed to be correlated with)
        
        Predictive: Can the measure predict something in the future that the construct should correlate wth
        
        Concurrent: Is the measure associated with performance on another task
        
        Assessing criterion validity requires the known-groups paradigm, presenting a new measure to groups of people whose state is known and seeing whether the measure can distinguish between them
      - Convergent: The measure is correlated with other measures that report the same construct
      - Discriminant: Does your measure NOT correlate with things it's NOT supposed to measure
- - - - Avoiding reactivity (observer effect) requires:
        
        Unobtrusive observation: staying hidden from whoever you're observing
        
        Habituation: People being observed get used to your presence
        
        Measure behavioural traces instead
        
        Use archival data instead