SensSpec.demo: Demonstrate Sensitivity, Specificity, PPV, and NPV
In TeachingDemos: Demonstrations for Teaching and Learning

Description Usage Arguments Details Value Author(s) See Also Examples

This function demonstrates how to get PPV and NPV from Sensitivity, Specificity, and Prevalence by using a virtual population rather than a direct application of Bayes Rule. This approach is more intuitive to mathphobes.

1	SensSpec.demo(sens, spec, prev, n = 100000, step = 11)

`sens`	Sensitivity (between 0 and 1)
`spec`	Specificity (between 0 and 1)
`prev`	Prevalence (between 0 and 1)
`n`	Size of the virtual population (large round number)
`step`	which step of the process to display

The common way to compute Positive Predictive Value (probability of disease given a positive test (PPV)) and Negative Predictive Value (probability of no disease given negative test (NPV)) is to use Bayes' rule with the Sensitivity, Specificity, and Prevalence.

This approach can be overwhelming to non-math types, so this demonstration goes through the steps of assuming a virtual population, then filling in a 2x2 table based on the population and given values of Sensitivity, Specificity, and Prevalence. PPV and NPV are then computed from this table. This approach is more intuitive to many people.

The function can be run multiple times with different values of step to show the steps in building the table, then rerun with different values to show how changes in the inputs affect the results.

An invisible matrix with the 2x2 table

Greg Snow, 538280@gmail.com

roc.demo, fagan.plot, the various Epi packages, tkexamp

for(i in seq(1,11,2)) {
  SensSpec.demo(sens=0.95, spec=0.99, prev=0.01, step=i)
  if( interactive() ) {
    readline("Press Enter to continue")
  }
}

          Disease
Test          Yes     No      Total
  Positive                         
  Negative                         
                                   
  Total                       1e+05
 
PPV =
NPV = 

          Disease
Test          Yes     No      Total
  Positive                         
  Negative                         
                                   
  Total      1000  99000      1e+05
 
PPV =
NPV = 

          Disease
Test          Yes     No      Total
  Positive    950                  
  Negative     50                  
                                   
  Total      1000  99000      1e+05
 
PPV =
NPV = 

          Disease
Test          Yes     No      Total
  Positive    950    990           
  Negative     50  98010           
                                   
  Total      1000  99000      1e+05
 
PPV =
NPV = 

          Disease
Test          Yes     No      Total
  Positive    950    990       1940
  Negative     50  98010      98060
                                   
  Total      1000  99000     100000
 
PPV =
NPV = 

          Disease
Test          Yes     No      Total
  Positive    950    990       1940
  Negative     50  98010      98060
                                   
  Total      1000  99000     100000
 
PPV = 950/1940 = 0.4897
NPV = 98010/98060 = 0.9995