PA Statistical Significance

A PA significance of 95% or higher, means that the calculated PA relevancy is reliable.
(This is all you really need to know. If you want the details, keep reading)

The significance of a test ranges from 0% to 100%. If can be described as the probability that a test result is 'real' and not just a coincidence. A significance above 95% is usually considered 'significant' by scientists and statisticians.


A word of warning





How is the significance calculated?

For PA Relevacies higher than 10%, I calculate the probablility (P) , that the score could occur by coincidence if PA:s (people who have prosopagnosia). and NT:s (people who don't have prosopagnosia) actually usually score the same.


The significance is based on 2 values:
X                The total number of testruns (counting only first time testruns by people who have clearly stated whether or not. they have prosopagnosia)
N                 The number of PA:s who "scored like NT:s" plus the number of NT:s who "scored like PA:s".

The probability (P) for the results (or more extreme results) is calculated like this:
P = ( 1 + X!/(X-1)!/1! + X!/(X-2)!/2! + X!/(X-3)!/3! + ... + X!/(X-N)!/N! ) / 2X

The significance in percent (%) is calculated like this:
Significance = ( 100 - 100 * P )



For PA Relevacies lower than 10%, I calculate the probablility (P) , that the score could be this even by coincidence, if PA:s actually usually score 10% higher (or lower) than NT:s (which equals a PA relevancy: of 20%).


The significance is based on 3 values:
X                The total number of testruns (counting only first time testruns by people who have clearly stated whether or not. they have prosopagnosia)
N                 The number of PA:s who "scored like NT:s" plus the number of NT:s who "scored like PA:s".
Q                 The probablility of a PA to "score like NT:s" or of an NT to "score like PA:s" if the real PA Relevancy were 20%. (=0.45)

The probability (P) for these results (or more even results) is calculated like this:
P = QN*(1-Q)(X-N)*X!/(X-N)!/N! + QN+1*(1-Q)(X-(N+1))*X!/(X-(N+1))!/(N+1)! + QN+2*(1-Q)(X-(N+2))*X!/(X-(N+2))!/(N+2)! + ... + QX-N*(1-Q)(X-(X-N))*X!/(X-(X-N))!/(X-N)!

The significance in percent (%) is calculated like this:
Significance = ( 100 - 100 * P )




The significance will not be calculated for a test until the PA relevancy has been calculated.








Home / Face-Blindness