Probability plots based on Student's t ‐distribution

The validity of the normal distribution as an error model is commonly tested with a (half) normal probability plot. Real data often contain outliers. The use of t ‐distributions in a probability plot to model such data more realistically is described. It is shown how a suitable value of the parameter ν of the t ‐distribution can be determined from the data. The results suggest that even data that seem to be modeled well using a normal distribution can be better modeled using a t ‐distribution.