View Single Post
Old 2022-09-25, 17:54   #5
Batalov
 
Batalov's Avatar
 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

5×1,997 Posts
Default

Quote:
Originally Posted by sticky View Post
What I dont understand is " graph is drawn in a sigma-scale for the y-axis"".
It is easy to understand by analogy. Suppose you have some variable that has an exponential behavior (for example any biological effect that is measured by luminescence, e.g. gene expression). To make it look linear, you change the plot to have y-axis transformed by the function inverse to exponent. What function is this? It is log().

Same here. CDF is known to have a wave-like, monotone function, usually denoted 𝛷() that describes it (when underlying distribution is normal). How do you transform y-axis? You apply the inverse function 𝛷-1(). It doesn't have an analytic expression but that doesn't matter - you can use a function that is indistinguishable from precise function, within pixels on your screen - you will not know the difference. So that's what programs like SigmaPlot do. It is rather simple.
Batalov is offline   Reply With Quote