LabVIEW
Probability plot - question
The probability form of the y-axis in the minitab example is often called a "probit scale". JMP also has capability to format cumulative data on a probit scale, for example in the following plot showing a weighted regression of data that are predominantly log-normal, but with tails on the upper and lower ends of the distribution, where the weight factors associated with each point are represented by the color.
An alternate approach is to convert the cumulative data (cumulative distribution function = cdf) to a cumulative probability function (cpf), for example...
cpf(x)=√2×inverf(2×cdf(x)-1)
... where inverf is the inverse error function, included in LabView Math Functions, and x=log(size) in this example.
The transformed plot, with JMP's weighted linear fits, looks like...
... where the y-scale represents the number of standard deviations from the mean (0 = mean, +1 = 84th percentile, +2 = 97th percentile, etc...). In LV, it looks like...
... where the "Linear model" is the cpf.
It would be nice to have the probit scale option in LV, enabling one to plot the cdf directly.
