On a probability plot the ordered data are drawn against a nonlinear probability axis chosen so that a sample from the reference distribution dist falls approximately along a straight line. Systematic departures from the reference line indicate departures from the distribution.

dist is one of "normal" (the default when dist is omitted), "lognormal", "exponential", "extreme value", "weibull", "rayleigh", "logistic", or "loglogistic". For "lognormal", "weibull", and "loglogistic" the data axis is logarithmic.

y is a numeric vector, or a matrix in which case each column is plotted as a separate sample. NaN values are ignored.

cens is a logical vector the same size as y that is true for right-censored observations; censored points are not plotted and the plotting positions of the remaining points follow the Kaplan-Meier estimate. freq is a vector of nonnegative integer frequencies (counts) the same size as y. Pass [] to omit either one.

probplot (ax, …) plots into the axes ax instead of the current axes. The trailing option "noref" suppresses the reference line.

h = probplot (…) returns a column vector of handles to the plotted line objects (the data markers, followed by the reference line unless "noref" was given).

The reference line is a robust fit through the first and third quartiles of the data on the transformed scale. For censored data the quartiles are taken from the Kaplan-Meier plotting positions; when heavy censoring prevents the data from reaching a quartile the position is linearly extrapolated, which may deviate slightly from MATLAB.

See also: normplot, wblplot, qqplot, cdfplot, ecdf

Source Code: probplot