Kurtosis is a measure of how sharp the data peak is. Traditionally the value of this coefficient is compared to a value of 0.0, which is the coefficient of kurtosis for a normal distribution (i.e. the bell-shaped curve). A value greater than 0 indicates a peaked distribution and a value less than 0 indicates a flat distribution. Without a very large sample size, the use of this coefficient is of questionable value.
Population Kurtosis ( )
Sample Kurtosis ( )
(adapted from King and Julstrom, 1982)
where
s |
|
N |
= number of data values for a population |
n |
= number of data values for a sample |
xi |
= ith data value |
See Also
95% and 99% Confidence Interval for the Mean
Critical Value of K-S Statistic at 90%, 95%, and 99% Significance Level
Kolmogorov-Smirnov Goodness of Fit Statistic for Normal Distribution