statistics.txt

  1. Sensitivity
    TP/(TP+FV)

    Only includes numbers about subjects with disease

    Sensitivity rules out (SNOUT)
  2. Specificity
    TN/(TN+FP)

    Only includes numbers on subjects without disease

    Specificity rules in (SPIN)
  3. Draw the bayesian foursquare
    Image Upload 2
  4. Positive Predictive Value
    Likelihood of disease in subject with positive test

    PPV=TP/(TP+FP)

    • depends on prevalence
    • (if uncommon, number of FP increases and PPV decreases)
    • good PPV means good screening
  5. Negative Predictive Value
    Likelihood of not having disease if tested negative

    NPV=TN/(TN+FN)

    • Depends on prevalence
    • (if common, number of FN increases and NPV decreases)
  6. type 1 error
    concluding difference when non exists (think drug companies)
  7. type 2 error
    concluding no difference when one does exist
  8. number needed to treat
    inverse of absolute risk reduction
Author
jmeglin
ID
27529
Card Set
statistics.txt
Description
Statistics
Updated