Excess kurtosis measures tail behavior relative to a normal distribution. A value of zero matches normal-distribution tails; positive values indicate more outliers, while negative values indicate fewer. section 12.5.4
Interpret the statistic by comparing its value with zero:
- Zero indicates tail behavior matching a normal, or Gaussian, distribution. section 12.5.4
- Positive values indicate more extreme observations or outliers than a typical normal distribution. section 12.5.4
- Negative values indicate fewer extreme observations than a typical normal distribution. section 12.5.4
Computing mantissa kurtosis
For population data transformed into mantissas, compute Fisher excess kurtosis with SciPy: `from scipy.stats import kurtosis; kurtosis(populations['mantissa'], fisher=True)`. The observed result is approximately `-1.178`, indicating fewer tails than a normal distribution. section 12.5.4
>>> from scipy.stats import kurtosis >>> print(kurtosis(populations['mantissa'], fisher = True)) -1.177858418956045
Inspect the direct SciPy calculation used to obtain the excess-kurtosis value. section 12.5.4
