Descriptive stats with SciPy

Beyond np.mean, stats.describe returns count, mean, variance, skewness, and kurtosis in one call—useful for quick EDA on ndarray exports from Pandas.

Key functions

• stats.describe — nobs, minmax, mean, variance, skewness, kurtosis
• stats.tstd, stats.tvar — sample std/var (ddof=1 default)
• stats.sem — standard error of the mean
• stats.iqr — robust spread measure
• stats.zscore — standardize for comparison

describe vs describe

Pandas df.describe() is tabular and column-wise. stats.describe targets one numeric array—ideal after series.to_numpy().

Example

import numpy as np
from scipy import stats

x = np.array([10, 12, 11, 15, 9, 14, 13])
d = stats.describe(x)
print('nobs:', d.nobs, 'mean:', d.mean, 'var:', d.variance)

Important interview questions and answers

1. Q: skewness meaning?
   A: Asymmetry: positive skew = long right tail; affects choice of mean vs median.
2. Q: zscore use?
   A: Compare values on different scales; watch outliers—zscore is sensitive.

Self-check

1. What five quantities does stats.describe return?
2. When prefer median and IQR over mean and std?

Tip: Pair stats.describe with a histogram mentally—skewness without a plot misleads.