Please enable JavaScript.

Coggle requires JavaScript to display documents.

Psy 350, EXAM 4, Exam 2, Exam 3, construct validity is very important for…

- - - - construct
      - statistical
        
        MARGIN OF ERROR
        
        confidence intervals
      - external
    - - statistical
        
        confidence interval, strength ofmcorrelation
      - external
      - construct
    - - statistical
      - external
      - construct
      - internal
  - - - how well did we operationalize/measure/manipulate variable(s)
      - ways to operationalize
        
        self report
        
        observational
        
        physiological
      - scales of measurement
        
        quantitative
        
        interval
        
        ratio
        
        there is a 0 (age)
        
        ordinal
        
        categorical
      - to assess
        
        quantitative
        
        discriminant
        
        convergent
        
        criterion (predict behavior, known groups paradigm)
        
        subjective
        
        content validity (all aspects)
        
        face validity
    - - is the statistical conclusion accurate to reality
    - - are there possible 3rd variables
    - - we can we generalize this to
  - - - 2 variables, measured
    - - 2 variables, one manipulated
        
        temporal precednce
        
        internal validity (random assignment, no 3rd variables)
        
        covariance
    - - % or # result
      - one variable, measured
  - - - categorical: kappa >/= .7
      - quantitative: correlation >/= .7
    - - cronbach's alpha >/= .8
    - - r >/= .5
- - - - interval scale
      - ratio
      - ordinal scale
- - - - posttest only
      - pretest/postest
  - - - selection effects
      - design confounds
      - observer effects.bias
- - - - fidelity/responsibilty
      - integrity
    - - beneficence
        
        help people
        
        do no harm
      - respect for persons
      - justice
  - - - refine experiment to reduce distress
    - - few animals as possible
    - - use other than animals when we can
- - - - enhances int. validity
    - - addresses specific internal validity threats
- - - - preregistration
    - - open materials
    - - address: open data
- - - - cross-sectional
      - cross lag
      - autocorrections
  - - - criterion
      - predictor
    - - mediation
        
        mediation hypotheses: causal statementx
        
        give a reason why they are related
        
        specify time sequence (temporal precedence)
    - - predicts DV controlling for other factors
      - strength of relationship controlling for other variables
  - - - can we generalize?
      - random sample?
    - - CI
      - other studies
    - - we are only measuring, not manipulating so we have less control
      - cant rule out things we didn't study
      - longitude\nal only establishes temporal precendecn
- - - - quantitative
        
        scatterplot
      - categorical: bar graph
  - - - statistical validity
        
        strength of relationship
        
        effect size: strength of relationship
        
        stat sig? precision? CI?
        
        ci contain 0?
        
        replicated?
        
        outliers
        
        have a larger effect when smaller sample size
        
        restriction of range?
        
        focusing on only part of the whole range
        
        curvilinear?
      - internal validity
        
        we can't make a causal claim
        
        b/c directionality not established and their variable problem
        
        third variable
        
        spurious association
      - construct validity
        
        ask about each variable
        
        face, content, criterion, convergent, discriminant
      - external validity
        
        how did they get the sample
        
        external validity does matter b/c we want to generalize
        
        but you can generalize even w/o random sampling
        
        Moderators
        
        outside variable tar changes the relationship b/n first 2 variables (eg gender)
- - - - double barreled
      - negatively worded
      - leading questions
    - - construct validity
        
        threatened by shortcuts
        
        fence sitting
        
        response sets
        
        acquiescence
        
        threatened by biases
        
        faking bad
        
        implicit associations (IAT)
        
        socially desirable responding
        
        how to fix:
        
        q's to catch socially desirable responding
        
        reverse wiring and ask question
        
        IAT
        
        remind anonymity
        
        discourages shortcuts
        
        remove neutral option (fence sitting)
        
        change q order
    - - forced choice
      - Llikert scale (Likert type scale)
      - open ended
      - semantic differential format
  - - - observer effects
        
        when observer's expectation accidentally changes participants behavior to match the expectation
        
        aka expectancy effects
      - observer bias
        
        in the observer
        
        when the observer sees what they want to see
      - reactivity
        
        when particpant changes behavior because of the present of observer
        
        solutions:
        
        unobtrusive observations
        
        wait it out: participants get used to your presence
        
        measure behavior RESULTS
      - preventing observer effects and bias
        
        masked design/ blind design
        
        codebook (interrater)
- - - - larger samples give smaller margin of error
      - random sampling
  - - - purposive
      - convenience
        
        self selection
      - snowballing
      - quota
    - - cluster
      - multistage
      - systematic
      - stratified random
      - simple random sampling
      - oversampling
- - - - concurrent
      - repeated
    - - pre test/ post test
      - post test only
    - - 1 group pretest posttest
    - - within
        
        pro
        
        fewer people
        
        everyone is their own control
        
        con
        
        order effects (beware of practice effects)
        
        practicality
        
        demand charcteristics
      - independent
        
        con
        
        selection effects (beware of individual differences)
        
        need more people
        
        pro
        
        no order effects risks
  - - - ruling out alternative explanations
      - IF WELL DESIGNED
        
        3 threats
        
        selection effects
        
        solution:
        
        random assignmnet
        
        matched asisgnent
        
        Independent groups
        
        order effects
        
        Within groups
        
        solution
        
        counterbalancing
        
        (only when 3 or fewer variables)
        
        partial counterbalancing
        
        more than 3 variables
        
        1 more item...
        
        2 kinds
        
        carryover effects (contamination)
        
        practice effects
        
        design confounds
        
        systematic variability
        
        solution
        
        control othewr variables
  - - - how precise was the estimate
    - - can we rule out any alternative explanations?
    - - was there random sampling
    - - was there a manipulation check?
- - - - ceiling/ floor effects (too easy or too hard)
      - insensitive measures
        
        refine the scale
      - solution: manipulation check
      - weak manipulation
    - - individual differences
        
        matched groups
      - situation noise
        
        control environment
      - measurement error
        
        larger sample size
  - - - attrition
      - testing
      - regression
      - instrumentation
      - history
      - maturation
    - - observer bias
        
        double blind study
      - demand characteristics
        
        double blind
        
        masked study
      - placebo effect
        
        double blind placebo controlled study
    - - selection effects
      - order effects
      - design confounds