Standard deviation describes how widely numeric values vary around their mean. In a dataset, it provides a quick sense of whether a feature is nearly constant or meaningfully dispersed. section 2.4.2
Using dispersion to screen features
For the Auto MPG numeric features, standard deviations include 7.82 for mpg, 105.21 for displacement, 38.52 for horsepower, 849.83 for weight, and 2.82 for acceleration. A feature with zero or very low variance may be excluded because it adds little information, provided missingness is considered separately. section 2.4.2
data[numeric_feats].std()
Compute dispersion for selected numeric columns with data[numeric_feats].std(). section 2.4.2
Standard deviation also supports comparison after standardization: subtract each feature’s mean and divide by its standard deviation before inspecting boxplots. section 2.4.2
