Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Exploratory multivariate technique for contingency tables. Decomposes the chi-squared distance between rows and columns of a frequency table into principal axes:
$$
\chi^2 = N \sum_{ij} \frac{(p_{ij} - p_{i+}p_{+j})^2}{p_{i+}p_{+j}}
$$
Biplot shows rows (types) and columns (attributes) in low-dimensional space, revealing typological patterns.
Step 3: Lithic Analysis — Reduction Sequence and Debitage
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.stats import spearmanr, kruskal
# ------------------------------------------------------------------ ## Lithic assemblage analysis: reduction sequence from debitage# ------------------------------------------------------------------ #
np.random.seed(42)
n_lithics = 400# Flake attributes (simulated)
lithics_data = {
'artifact_id': [f"L{i:04d}"for i inrange(n_lithics)],
'site': np.random.choice(['Area 1', 'Area 2', 'Area 3'], n_lithics, p=[0.5, 0.3, 0.2]),
'material': np.random.choice(['Flint', 'Obsidian', 'Chert', 'Quartzite'], n_lithics,
p=[0.50, 0.20, 0.25, 0.05]),
'type': np.random.choice(['Primary flake', 'Secondary flake', 'Interior flake',
'Blade', 'Core', 'Biface', 'Tool'], n_lithics,
p=[0.15, 0.25, 0.30, 0.10, 0.05, 0.05, 0.10]),
'cortex_pct': np.random.choice([0, 10, 25, 50, 75, 100], n_lithics),
'length_mm': np.random.lognormal(3.2, 0.5, n_lithics),
'width_mm': np.random.lognormal(2.9, 0.5, n_lithics),
'thickness_mm': np.random.lognormal(1.8, 0.5, n_lithics),
'platform': np.random.choice(['cortical', 'plain', 'faceted', 'punctiform', 'absent'], n_lithics,
p=[0.15, 0.40, 0.25, 0.10, 0.10]),
'flake_scars': np.random.randint(0, 15, n_lithics),
'retouched': np.random.choice([True, False], n_lithics, p=[0.15, 0.85]),
}
lithics_df = pd.DataFrame(lithics_data)
# Compute reduction index (proxy: log weight ÷ dorsal scar count)
lithics_df['weight_g'] = (lithics_df['length_mm'] * lithics_df['width_mm'] * lithics_df['thickness_mm']) * 0.002
lithics_df['reduction_index'] = (lithics_df['flake_scars'] + 1) / (lithics_df['cortex_pct']/100 + 0.1)
print(f"Lithic assemblage: {len(lithics_df)} artifacts")
print("\nType distribution:")
print(lithics_df['type'].value_counts())
# ---- Reduction stage by cortex percentage ---------------------- #
cortex_by_type = lithics_df.groupby('type')['cortex_pct'].describe()
print("\nCortex % by artifact type:")
print(cortex_by_type[['mean','std','50%']].round(1))
# ---- Platform × flake type association ------------------------- #
ct_plat = pd.crosstab(lithics_df['type'], lithics_df['platform'])
from scipy.stats import chi2_contingency
chi2, p, _, _ = chi2_contingency(ct_plat)
print(f"\nPlatform × type: χ²={chi2:.2f}, p={p:.4f}")
# ---- Metric analysis by material -------------------------------- ## Kruskal-Wallis test for length differences by material
materials = lithics_df['material'].unique()
groups = [lithics_df[lithics_df['material']==m]['length_mm'].values for m in materials]
h_stat, p_kw = kruskal(*groups)
print(f"\nLength by material (Kruskal-Wallis): H={h_stat:.2f}, p={p_kw:.4f}")
# ---- Visualization --------------------------------------------- #
fig, axes = plt.subplots(2, 2, figsize=(13, 9))
# Type distribution
type_counts = lithics_df['type'].value_counts()
colors_lit = ['#8B4513','#D2691E','#A0522D','#CD853F','#DEB887','#F4A460','#FFDEAD']
axes[0][0].barh(type_counts.index, type_counts.values, color=colors_lit[:len(type_counts)],
edgecolor='black', linewidth=0.7)
axes[0][0].set_xlabel("Count"); axes[0][0].set_title("Lithic Type Distribution")
axes[0][0].grid(axis='x', alpha=0.3)
# Reduction sequence: cortex by type
lithics_df.boxplot(column='cortex_pct', by='type', ax=axes[0][1])
axes[0][1].set_title("Cortex % by Artifact Type (Reduction Stage)")
axes[0][1].set_xlabel("Artifact type"); axes[0][1].set_ylabel("Cortex %")
plt.sca(axes[0][1]); plt.xticks(rotation=30, fontsize=7)
# Length distribution by materialfor mat in materials:
subset = lithics_df[lithics_df['material']==mat]['length_mm']
axes[1][0].hist(subset, bins=20, alpha=0.5, density=True, label=mat)
axes[1][0].set_xlabel("Length (mm)"); axes[1][0].set_ylabel("Density")
axes[1][0].set_title("Flake Length Distribution by Raw Material")
axes[1][0].legend(fontsize=8); axes[1][0].grid(True, alpha=0.3)
# Length × width scatter (flake shape)
axes[1][1].scatter(lithics_df['length_mm'], lithics_df['width_mm'],
c=['#3498db'if r else'#e74c3c'for r in lithics_df['retouched']],
s=15, alpha=0.5)
axes[1][1].set_xlabel("Length (mm)"); axes[1][1].set_ylabel("Width (mm)")
axes[1][1].set_title("Flake Dimensions (blue=unretouched, red=retouched)")
axes[1][1].grid(True, alpha=0.3)
plt.tight_layout()
plt.savefig("lithic_analysis.png", dpi=150)
plt.show()
Advanced Usage
Hierarchical Cluster Analysis of Assemblages
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from scipy.cluster.hierarchy import dendrogram, linkage
from scipy.spatial.distance import pdist
# Build site-level attribute matrix (ware proportions)
ct_site_ware = pd.crosstab(df['site'], df['ware'], normalize='index')
# Bray-Curtis dissimilarity (for compositional data)try:
from scipy.spatial.distance import braycurtis
bc_matrix = np.zeros((len(ct_site_ware), len(ct_site_ware)))
for i inrange(len(ct_site_ware)):
for j inrange(len(ct_site_ware)):
bc_matrix[i,j] = braycurtis(ct_site_ware.iloc[i], ct_site_ware.iloc[j])
except Exception:
bc_matrix = pdist(ct_site_ware.values, metric='euclidean')
from scipy.spatial.distance import squareform
try:
dist_condensed = squareform(bc_matrix)
except Exception:
dist_condensed = pdist(ct_site_ware.values, metric='euclidean')
Z = linkage(dist_condensed, method='ward')
fig, ax = plt.subplots(figsize=(8, 5))
dendrogram(Z, labels=ct_site_ware.index.tolist(), ax=ax)
ax.set_title("Assemblage Cluster Analysis (Ward linkage, Bray-Curtis dissimilarity)")
ax.set_xlabel("Site"); ax.set_ylabel("Dissimilarity")
ax.grid(axis='y', alpha=0.3)
plt.tight_layout(); plt.savefig("assemblage_cluster.png", dpi=150); plt.show()
Troubleshooting
CA axes flip sign unexpectedly
SVD sign is arbitrary. Multiply both row and column coordinates by -1 for a given axis if needed for interpretability — the relative positions remain the same.
Chi-square test invalid with small expected frequencies
Fix: Merge rare categories before testing:
# Merge rare wares (<10 counts) into "Other"
rare = ware_counts[ware_counts < 10].index
df['ware_merged'] = df['ware'].replace({w: 'Other'for w in rare})
Outlier sherds distort CA
Fix: Remove outlier rows/columns with very low frequencies before CA:
ct_filtered = ct_ca[ct_ca.sum(axis=1) >= 5] # Min 5 sherds per context
ct_filtered = ct_filtered.loc[:, ct_filtered.sum(axis=0) >= 5]
