Applying nonparametric tests in experimental analysis

Experimental Design in Python

James Chapman

Curriculum Manager, DataCamp

When to use nonparametric tests

Parametric test assumptions not met
Data on ordinal scale or non-normal
Robust to outliers and non-linear data

Skewed distribution

Exploring nonparametric methods

Nonparametric methods for data that doesn't fit traditional assumptions
Mann-Whitney U Test: compare two independent groups
Kruskal-Wallis Test: compare more than two groups

Non-normal distribution

Visualizing nonparametric data

Violin plots: Visualizing distributions across groups

condensed_data = mineral_rocks[mineral_rocks['RockType'].isin(['Igneous', 'Metamorphic'])]

sns.violinplot(x='RockType', y='MineralHardness', data=condensed_data)

Violin plot of Mineral Hardness by Rock Type (Igneous vs Metamorphic)

Visualizing nonparametric data

Boxen plot: better insight into distribution shape

sns.boxenplot(x='RockType', y='MineralHardness', data=mineral_rocks)

Boxen plot of Mineral Hardness Distribution Across Rock Types

Applying nonparametric tests - Mann Whitney U

from scipy.stats import mannwhitneyu, kruskal
u_stat, u_pval = mannwhitneyu(
    mineral_rocks[mineral_rocks['RockType'] == 'Igneous']['MineralHardness'],
    mineral_rocks[mineral_rocks['RockType'] == 'Sedimentary']['MineralHardness']
)

print(f"Mann-Whitney U test p-value: {u_pval:.4f}")

Mann-Whitney U test p-value: 0.9724

Applying nonparametric tests - Kruskal-Wallis

k_stat, k_pval = kruskal(
    mineral_rocks[mineral_rocks['RockType'] == 'Igneous']['MineralHardness'],
    mineral_rocks[mineral_rocks['RockType'] == 'Sedimentary']['MineralHardness'],
    mineral_rocks[mineral_rocks['RockType'] == 'Metamorphic']['MineralHardness']
)

print(f"Kruskal-Wallis test p-value: {k_pval:.4f}")

Kruskal-Wallis test p-value: 0.0630

Let's practice!

Experimental Design in Python