Please enable JavaScript.

Coggle requires JavaScript to display documents.

TASK 9 (Dawes (Actuarial/statistical method: human judge is eliminated and…

- - - - Both methods should base judgements on the same data
      - One must avoid conditions that can artificially inflate the accuracy of actuarial methods
      - o Cross-validation needed  allows to determine how the method performs realistically
    - - 1. Judgments mediated by theories  difficult/ impossible to duplicate by statistical frequencies alone
      - Select reversal of actuarial conclusions based on the consideration of rare events or utility functions that are not incorporated into statistical methods
        o ‘Broken leg’ problem
      - Complex configural relations between predictive variables and outcome
        o Only human observer may notice a particular facial expression or mannerism that has true predictive value
    - - When properly derived, the mathematical features of actuarial methods ensure that variables contribute to conclusions based on their actual predictive power and relation to the criterion of interest
      - random fluctuations in humans which decrease reliability and accuracy
        Clinical judgements can produce ‘self-fulfilling prophecies’
      - Optimal weights: of variables is less important in actuarial method than commonly assumed
      - Lack of impact: research on the differences between methods had little impact on everyday decision making
- - - - In my judgement, can this decision problem be decomposed into simple segments?
      - What are the consequences of alternative actions of the decision?
      - What are the uncertainties in the environment relevant to the actions and their consequences
      - If one considers all 3 options, the clinical (head), the actuarial (formula) or the combination, the accuracy of the actuarial method equals/ exceeds the clinical method in the literature
        
        but Research on comparison between actuarial and combination also favours the actuarial alone
        
        Question of generalisation to other fields: they agree with Kleinmuntz that humans show certain unique capabilities (e.g. visual pattern analysis)  can provide potential useful input for decision purposes
        
        - Still, a clinician can then record the impressions and form an actuarial analysis
- - - - o The lower the prevalence, the worse the PPV and the better the NPV
      - low prevalance = more false positives
    - - calculated from specificity and sensitivity
      - not affected by prevalance
- - - - o Rule of thumb: high prevalence in population/ large benefit from finding true cases  lenient threshold, low prevalence/ high cost of false alarms  strict threshold
      -  Statistical prediction rules can often raise accuracy of repetitive diagnostic decisions and formulas for setting decision thresholds can improve the utility of those decisions
         Standardisation speeds up the diagnostic process, give decision makers a way to communicate more easily and precisely about impressionistic features
- - - - clinicians only look for what symptoms are present and which symptoms are present but dont fit, but dont take inco account symptoms that are NOT PRESENT BUT SHOULD IN A GIVEN CONDITION or CANT BE PRESENT IN A GIVEN CONDITION
    - - Implicit personality theory: most people think aggressive people are not likely to be friendly  inferences about other traits from the one observed
      - when drawings were randomly paired with personality traits presumably characteristic of the person who did the drawing, subjects fabricated illusory correlations between drawing features and personality traits
    - - Treatment effects: as long as a group/ patient improves in performance for any reason, the therapist may attribute the improvement to the treatment administered (Hawthorne effect)
      - Confirmation bias: people selectively seek evidence that confirms their hypothesis - People disregard information that contradicts their current judgement
      - Providing more information to a judge as found to increase their confidence in the decision without necessarily increasing the accuracy of the decision
    - - REDUCE
        
        Present people with 2 alternative questions  have to make pro and con list for each of the 2 possible options  then they chose one of the answers and stated confidence in correctness of choice (Koriat et al.) (eliminate overconfidence
        
        Reduce hindsight bias: ask subjects to explain why outcome A might have been expected from prior events and how alternative outcome B might be explained if it had occurred instead of A  forcing one to consider B reduced the hindsight bias toward A
        
        Reduce biases by entertaining alternative hypotheses for a long period of time  active consideration of alternative
  - - - Posterior odds: p that H is true given the piece of information divided by p that H is not true given the piece of information
    - - If theprior offs are x/y, the likelihood ratio must be larger than y/x in order for the hypothesis to be correct more than 50% of time
      - The more unlikely the hypothesis (the lower the prior odds), the greater the likelihood ratio must be to justify the hypothesis
- - - - o These are mutually exclusive and exhaustive classes of ways to combine data; the relative value of these 2 classes is a meaningful question, whether the data to be combined come from interviews, Rorschach protocols, Minnesota Multiphasic Personality Inventory-2 (MMPI-2) profiles, or behaviour counts
    - - o Clinician can be given the output of a statistical formula or a formula can include a variable representing quantified clinician judgement  the final prediction depends on trained judgement in former situation and not in latter
      - o Subtypes do not make distinction meaningless or arbitrary