Skip to content

K-Dense-AI/scientific-agent-skills / skills/seaborn

seaborn

Verified

Creates Seaborn statistical visualizations with pandas integration for distributions, relationships, categorical comparisons, regression displays, pair plots, and heatmaps. Supports function and objects interfaces with explicit aggregation, uncertainty, and missing-data handling. Best suited to static exploratory plots; plotly covers interactive figures and scientific-visualization covers publication styling.

references/examples.md

View on GitHub
# Seaborn Common Use Cases and Examples

Seaborn 0.13.2 recipes reviewed 2026-10-01. These are illustrative templates: supply the named DataFrames and adapt the sampling design before using the statistics. Corrected contracts were exercised with tiny synthetic fixtures; notebook widget display and arbitrary scientific datasets were not validated. Import NumPy, pandas, Matplotlib, and Seaborn as below before using later snippets.

## Exploratory Data Analysis

### Quick Dataset Overview

```python
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd

# Load data
df = pd.read_csv('data.csv')

# Pairwise relationships for all numeric variables
sns.pairplot(df, hue='target_variable', corner=True, diag_kind='kde')
plt.suptitle('Dataset Overview', y=1.01)
plt.savefig('overview.png', dpi=300, bbox_inches='tight')
```

### Distribution Exploration

```python
# Multiple distributions across categories
g = sns.displot(
    data=df,
    x='measurement',
    hue='condition',
    col='timepoint',
    kind='kde',
    fill=True,
    height=3,
    aspect=1.5,
    col_wrap=3,
    common_norm=False
)
g.set_axis_labels('Measurement Value', 'Density')
g.set_titles('{col_name}')
```

### Correlation Analysis

```python
# Compute correlation matrix
corr = df.select_dtypes(include='number').corr()

# Create mask for upper triangle
mask = np.triu(np.ones_like(corr, dtype=bool))

# Plot heatmap
fig, ax = plt.subplots(figsize=(10, 8))
sns.heatmap(
    corr,
    mask=mask,
    annot=True,
    fmt='.2f',
    cmap='coolwarm',
    center=0,
    square=True,
    linewidths=1,
    cbar_kws={'shrink': 0.8}
)
plt.title('Correlation Matrix')
plt.tight_layout()
```

## Scientific Publications

### Multi-Panel Figure with Different Plot Types

```python
# Set publication style
sns.set_theme(style='ticks', context='paper', font_scale=1.1)
sns.set_palette('colorblind')

# Create figure with custom layout
fig = plt.figure(figsize=(12, 8))
gs = fig.add_gridspec(2, 3, hspace=0.3, wspace=0.3)

# Panel A: Time series
ax1 = fig.add_subplot(gs[0, :2])
sns.lineplot(
    data=timeseries_df,
    x='time',
    y='expression',
    hue='gene',
    style='treatment',
    markers=True,
    dashes=False,
    ax=ax1
)
ax1.set_title('A. Gene Expression Over Time', loc='left', fontweight='bold')
ax1.set_xlabel('Time (hours)')
ax1.set_ylabel('Expression Level (AU)')

# Panel B: Distribution comparison
ax2 = fig.add_subplot(gs[0, 2])
sns.violinplot(
    data=expression_df,
    x='treatment',
    y='expression',
    inner='box',
    ax=ax2
)
ax2.set_title('B. Expression Distribution', loc='left', fontweight='bold')
ax2.set_xlabel('Treatment')
ax2.set_ylabel('')

# Panel C: Correlation
ax3 = fig.add_subplot(gs[1, 0])
sns.scatterplot(
    data=correlation_df,
    x='gene1',
    y='gene2',
    hue='cell_type',
    alpha=0.6,
    ax=ax3
)
sns.regplot(
    data=correlation_df,
    x='gene1',
    y='gene2',
    scatter=False,
    color='black',
    ax=ax3
)
ax3.set_title('C. Gene Correlation', loc='left', fontweight='bold')
ax3.set_xlabel('Gene 1 Expression')
ax3.set_ylabel('Gene 2 Expression')

# Panel D: Heatmap
ax4 = fig.add_subplot(gs[1, 1:])
sns.heatmap(
    sample_matrix,
    cmap='RdBu_r',
    center=0,
    annot=True,
    fmt='.1f',
    cbar_kws={'label': 'Log2 Fold Change'},
    ax=ax4
)
ax4.set_title('D. Treatment Effects', loc='left', fontweight='bold')
ax4.set_xlabel('Sample')
ax4.set_ylabel('Gene')

# Clean up
sns.despine()
plt.savefig('figure.pdf', dpi=300, bbox_inches='tight')
plt.savefig('figure.png', dpi=300, bbox_inches='tight')
```

### Box Plot with Significance Annotations

```python
import numpy as np
from scipy import stats
from statsmodels.stats.multitest import multipletests

# Assumes independent experimental units and two prespecified contrasts.
# Paired or clustered designs need a corresponding analysis.
comparisons = [('Control', 'High'), ('Control', 'Low')]
raw_p = []
for left, right in comparisons:
    a = df.loc[df['treatment'].eq(left), 'response'].dropna()
    b = df.loc[df['treatment'].eq(right), 'response'].dropna()
    if min(len(a), len(b)) < 2 or not np.isfinite(np.r_[a, b]).all():
        raise ValueError('Need at least two finite independent measurements per group')
    raw_p.append(stats.ttest_ind(a, b, equal_var=False).pvalue)
if not np.isfinite(raw_p).all():
    raise ValueError('A contrast was undefined; do not label it nonsignificant')
adjusted_p = multipletests(raw_p, method='holm')[1]

# Create plot
fig, ax = plt.subplots(figsize=(8, 6))
sns.boxplot(
    data=df,
    x='treatment',
    y='response',
    hue='treatment',
    order=['Control', 'Low', 'Medium', 'High'],
    palette='Set2',
    legend=False,
    ax=ax
)

# Add individual points
sns.stripplot(
    data=df,
    x='treatment',
    y='response',
    order=['Control', 'Low', 'Medium', 'High'],
    color='black',
    alpha=0.3,
    size=3,
    ax=ax
)

# Add significance bars
def add_significance_bar(ax, x1, x2, y, h, text):
    ax.plot([x1, x1, x2, x2], [y, y+h, y+h, y], 'k-', lw=1.5)
    ax.text((x1+x2)/2, y+h, text, ha='center', va='bottom')

y_max = df['response'].max()
add_significance_bar(ax, 0, 3, y_max + 1, 0.5, f'Holm p={adjusted_p[0]:.3g}')
add_significance_bar(ax, 0, 1, y_max + 3, 0.5, f'Holm p={adjusted_p[1]:.3g}')

ax.set_ylabel('Response (μM)')
ax.set_xlabel('Treatment Condition')
ax.set_title('Treatment Response Analysis')
sns.despine()
```

## Time Series Analysis

These aggregate examples assume independent replicates at each timestamp. Inspect missingness and replicate counts before plotting. `lineplot` bridges missing rows; use the explicit segment recipe in [patterns_and_troubleshooting.md](patterns_and_troubleshooting.md) for gaps. Autocorrelated samples are not independent replicates.

### Multiple Time Series with Confidence Bands

```python
# Plot with automatic aggregation
fig, ax = plt.subplots(figsize=(10, 6))
sns.lineplot(
    data=timeseries_df,
    x='timestamp',
    y='value',
    hue='sensor',
    style='location',
    markers=True,
    dashes=False,
    errorbar=('ci', 95),
    ax=ax
)

# Customize
ax.set_xlabel('Date')
ax.set_ylabel('Measurement (units)')
ax.set_title('Sensor Measurements Over Time')
ax.legend(title='Sensor & Location', bbox_to_anchor=(1.05, 1), loc='upper left')

# Format x-axis for dates
import matplotlib.dates as mdates
ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m-%d'))
ax.xaxis.set_major_locator(mdates.DayLocator(interval=7))
plt.xticks(rotation=45, ha='right')

plt.tight_layout()
```

### Faceted Time Series

```python
# Create faceted time series
g = sns.relplot(
    data=long_timeseries,
    x='date',
    y='measurement',
    hue='device',
    col='location',
    row='metric',
    kind='line',
    height=3,
    aspect=2,
    errorbar='sd',
    facet_kws={'sharex': True, 'sharey': False}
)

# Customize facet titles
g.set_titles('{row_name} - {col_name}')
g.set_axis_labels('Date', 'Value')

# Rotate x-axis labels
for ax in g.axes.flat:
    ax.tick_params(axis='x', rotation=45)

g.tight_layout()
```

## Categorical Comparisons

### Nested Categorical Variables

```python
# Create figure
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

# Left panel: Grouped bar plot
sns.barplot(
    data=df,
    x='category',
    y='value',
    hue='subcategory',
    errorbar=('ci', 95),
    capsize=0.1,
    ax=axes[0]
)
axes[0].set_title('Mean Values with 95% CI')
axes[0].set_ylabel('Value (units)')
axes[0].legend(title='Subcategory')

# Right panel: Strip + violin plot
sns.violinplot(
    data=df,
    x='category',
    y='value',
    hue='subcategory',
    inner=None,
    alpha=0.3,
    legend=False,
    ax=axes[1]
)
sns.stripplot(
    data=df,
    x='category',
    y='value',
    hue='subcategory',
    dodge=True,
    size=3,
    alpha=0.6,
    legend=False,
    ax=axes[1]
)
axes[1].set_title('Distribution of Individual Values')
axes[1].set_ylabel('')

plt.tight_layout()
```

### Point Plot for Trends

```python
# Show how values change across categories
sns.pointplot(
    data=df,
    x='timepoint',
    y='score',
    hue='treatment',
    markers=['o', 's', '^'],
    linestyles=['-', '--', '-.'],
    dodge=0.3,
    capsize=0.1,
    errorbar=('ci', 95)
)

plt.xlabel('Timepoint')
plt.ylabel('Performance Score')
plt.title('Treatment Effects Over Time')
plt.legend(title='Treatment', bbox_to_anchor=(1.05, 1), loc='upper left')
sns.despine()
plt.tight_layout()
```

## Regression and Relationships

### Linear Regression with Facets

```python
# Fit separate regressions for each category
g = sns.lmplot(
    data=df,
    x='predictor',
    y='response',
    hue='treatment',
    col='cell_line',
    height=4,
    aspect=1.2,
    scatter_kws={'alpha': 0.5, 's': 50},
    ci=95,
    palette='Set2'
)

g.set_axis_labels('Predictor Variable', 'Response Variable')
g.set_titles('{col_name}')
g.tight_layout()
```

### Polynomial Regression

```python
fig, axes = plt.subplots(1, 3, figsize=(15, 5))

for idx, order in enumerate([1, 2, 3]):
    sns.regplot(
        data=df,
        x='x',
        y='y',
        order=order,
        scatter_kws={'alpha': 0.5},
        line_kws={'color': 'red'},
        ci=95,
        ax=axes[idx]
    )
    axes[idx].set_title(f'Order {order} Polynomial Fit')
    axes[idx].set_xlabel('X Variable')
    axes[idx].set_ylabel('Y Variable')

plt.tight_layout()
```

### Residual Analysis

```python
fig, axes = plt.subplots(2, 2, figsize=(12, 10))

# Use identical complete finite observations in all diagnostics
df = df.loc[np.isfinite(df[['x', 'y']]).all(axis=1)].copy()
# Main regression
sns.regplot(data=df, x='x', y='y', ax=axes[0, 0])
axes[0, 0].set_title('Regression Fit')

# Residuals vs predictor
sns.residplot(data=df, x='x', y='y', lowess=True,
              scatter_kws={'alpha': 0.5},
              line_kws={'color': 'red', 'lw': 2},
              ax=axes[0, 1])
axes[0, 1].set_title('Residuals vs Predictor')
axes[0, 1].axhline(0, ls='--', color='gray')

# Q-Q plot (using scipy)
from scipy import stats as sp_stats
residuals = df['y'] - np.poly1d(np.polyfit(df['x'], df['y'], 1))(df['x'])
sp_stats.probplot(residuals, dist="norm", plot=axes[1, 0])
axes[1, 0].set_title('Q-Q Plot')

# Histogram of residuals
sns.histplot(residuals, kde=True, ax=axes[1, 1])
axes[1, 1].set_title('Residual Distribution')
axes[1, 1].set_xlabel('Residuals')

plt.tight_layout()
```

## Bivariate and Joint Distributions

### Joint Plot with Multiple Representations

```python
# With hue, scatter jointplot uses KDE marginals (not histogram kwargs).
g = sns.jointplot(
    data=df,
    x='var1',
    y='var2',
    hue='category',
    kind='scatter',
    height=8,
    ratio=4,
    space=0.1,
    joint_kws={'alpha': 0.5, 's': 50},
    marginal_kws={'fill': True, 'common_norm': False}
)

# Add reference lines
g.ax_joint.axline((0, 0), slope=1, color='r', ls='--', alpha=0.5, label='y=x')
g.ax_joint.legend()

g.set_axis_labels('Variable 1', 'Variable 2', fontsize=12)
```

### KDE Contour Plot

```python
fig, ax = plt.subplots(figsize=(8, 8))

# Bivariate KDE with filled contours
sns.kdeplot(
    data=df,
    x='x',
    y='y',
    fill=True,
    levels=10,
    cmap='viridis',
    thresh=0.05,
    ax=ax
)

# Overlay scatter
sns.scatterplot(
    data=df,
    x='x',
    y='y',
    color='white',
    edgecolor='black',
    s=50,
    alpha=0.6,
    ax=ax
)

ax.set_xlabel('X Variable')
ax.set_ylabel('Y Variable')
ax.set_title('Bivariate Distribution')
```

### Hexbin with Marginals

```python
# For large datasets
g = sns.jointplot(
    data=large_df,
    x='x',
    y='y',
    kind='hex',
    height=8,
    ratio=5,
    space=0.1,
    joint_kws={'gridsize': 30, 'cmap': 'viridis'},
    marginal_kws={'bins': 50, 'color': 'skyblue'}
)

g.set_axis_labels('X Variable', 'Y Variable')
```

## Matrix and Heatmap Visualizations

### Hierarchical Clustering Heatmap

```python
# Prepare data (samples x features)
data_matrix = df.set_index('sample_id')[feature_columns]

# Create color annotations
row_colors = df.set_index('sample_id')['condition'].map({
    'control': '#1f77b4',
    'treatment': '#ff7f0e'
})

col_colors = pd.Series(['#2ca02c' if 'gene' in col else '#d62728'
                        for col in data_matrix.columns], index=data_matrix.columns)

# Scaling rows compares each sample's across-feature profile. Use z_score=1
# only when standardizing each feature across samples matches the question.
if not data_matrix.index.is_unique or not data_matrix.columns.is_unique:
    raise ValueError('Sample and feature labels must be unique')
if not np.isfinite(data_matrix.to_numpy()).all() or (data_matrix.std(axis=1) == 0).any():
    raise ValueError('Clustering needs finite, nonconstant rows for z_score=0')
if row_colors.isna().any() or col_colors.isna().any():
    raise ValueError('Missing annotation color')

# Plot
g = sns.clustermap(
    data_matrix,
    method='ward',
    metric='euclidean',
    z_score=0,  # Normalize rows
    cmap='RdBu_r',
    center=0,
    row_colors=row_colors,
    col_colors=col_colors,
    figsize=(12, 10),
    dendrogram_ratio=(0.1, 0.1),
    cbar_pos=(0.02, 0.8, 0.03, 0.15),
    linewidths=0.5
)

g.ax_heatmap.set_xlabel('Features')
g.ax_heatmap.set_ylabel('Samples')
g.savefig('clustermap.png', dpi=300, bbox_inches='tight')
```

### Annotated Heatmap with Custom Colorbar

```python
# Pivot data for heatmap
pivot_data = df.pivot(index='row_var', columns='col_var', values='value')

# Create heatmap
fig, ax = plt.subplots(figsize=(10, 8))
sns.heatmap(
    pivot_data,
    annot=True,
    fmt='.1f',
    cmap='RdYlGn',
    center=pivot_data.mean().mean(),
    vmin=pivot_data.min().min(),
    vmax=pivot_data.max().max(),
    linewidths=0.5,
    linecolor='gray',
    cbar_kws={
        'label': 'Value (units)',
        'orientation': 'vertical',
        'shrink': 0.8,
        'aspect': 20
    },
    ax=ax
)

ax.set_title('Variable Relationships', fontsize=14, pad=20)
ax.set_xlabel('Column Variable', fontsize=12)
ax.set_ylabel('Row Variable', fontsize=12)

plt.xticks(rotation=45, ha='right')
plt.yticks(rotation=0)
plt.tight_layout()
```

## Statistical Comparisons

### Before/After Comparison

Retain complete matched subjects for a paired-effect analysis. The red timepoint intervals below describe marginal means, not uncertainty in the change. Analyze and plot `after - before` per subject when the treatment effect is the estimand.

```python
# This wide input has one row per subject.
if not df['subject'].is_unique:
    raise ValueError('Expected one before/after pair per subject')
# Select the paired columns so unrelated columns cannot collide with value_name.
df_paired = df[['subject', 'before', 'after']].melt(
    id_vars='subject',
    value_vars=['before', 'after'],
    var_name='timepoint',
    value_name='measurement'
)

fig, axes = plt.subplots(1, 2, figsize=(12, 5))

# Left: Individual trajectories
for subject in df_paired['subject'].unique():
    subject_data = df_paired[df_paired['subject'] == subject]
    axes[0].plot(subject_data['timepoint'], subject_data['measurement'],
                 'o-', alpha=0.3, color='gray')

sns.pointplot(
    data=df_paired,
    x='timepoint',
    y='measurement',
    color='red',
    markers='D',
    markersize=8,
    errorbar=('ci', 95),
    capsize=0.2,
    ax=axes[0]
)
axes[0].set_title('Individual Changes')
axes[0].set_ylabel('Measurement')

# Right: Distribution comparison
sns.violinplot(
    data=df_paired,
    x='timepoint',
    y='measurement',
    inner='box',
    ax=axes[1]
)
sns.swarmplot(
    data=df_paired,
    x='timepoint',
    y='measurement',
    color='black',
    alpha=0.5,
    size=3,
    ax=axes[1]
)
axes[1].set_title('Distribution Comparison')
axes[1].set_ylabel('')

plt.tight_layout()
```

### Dose-Response Curve

Use numeric concentration coordinates. An ordinal strip/point axis makes unequally
spaced doses look equally spaced and cannot support a numeric fitted curve.
This illustrative fit assumes a monotone increasing four-parameter response and
independent observations. A fit or small residual is not evidence of identifiable
EC50: inspect uncertainty, residuals, plate effects, and coverage of both plateaus.
Zero-dose controls need separate display/model handling, not a logarithm.

```python
from scipy.optimize import curve_fit
from scipy.special import expit

fit_data = dose_df[['dose', 'response']].astype(float).copy()
if not np.isfinite(fit_data.to_numpy()).all() or (fit_data['dose'] <= 0).any():
    raise ValueError('This log-dose recipe requires finite responses and positive doses')
if fit_data['dose'].nunique() < 5:
    raise ValueError('Use at least five distinct positive doses for this demonstration')

def sigmoid(x, bottom, amplitude, ec50, hill):
    return bottom + amplitude * expit(hill * (np.log(x) - np.log(ec50)))

x = fit_data['dose'].to_numpy()
y = fit_data['response'].to_numpy()
span = np.ptp(y)
if span == 0:
    raise ValueError('Constant response cannot identify a dose-response curve')
params, covariance = curve_fit(
    sigmoid, x, y, p0=[y.min(), span, np.median(x), 1.],
    bounds=([-np.inf, 0., np.finfo(float).tiny, .01],
            [np.inf, np.inf, np.inf, 10.]), maxfev=10000,
)
x_smooth = np.geomspace(x.min(), x.max(), 200)
fig, ax = plt.subplots(figsize=(8, 6))
sns.scatterplot(data=fit_data, x='dose', y='response', color='gray', alpha=.4, ax=ax)
sns.lineplot(data=fit_data, x='dose', y='response', errorbar=('ci', 95),
             seed=7, marker='o', color='blue', ax=ax, label='Mean and 95% CI')
ax.plot(x_smooth, sigmoid(x_smooth, *params), color='red', label='Illustrative fit')
ax.set(xscale='log', xlabel='Dose (concentration units)', ylabel='Response')
ax.legend()
sns.despine(ax=ax)
```

## Custom Styling

### Custom Color Palette from Hex Codes

```python
# Define custom palette
custom_palette = ['#E64B35', '#4DBBD5', '#00A087', '#3C5488', '#F39B7F']
sns.set_palette(custom_palette)

# Or use for specific plot
sns.scatterplot(
    data=df,
    x='x',
    y='y',
    hue='category',
    palette=custom_palette
)
```

### Publication-Ready Theme

```python
# Set comprehensive theme
sns.set_theme(
    context='paper',
    style='ticks',
    palette='colorblind',
    font='Arial',
    font_scale=1.1,
    rc={
        'figure.dpi': 300,
        'savefig.dpi': 300,
        'savefig.format': 'pdf',
        'axes.linewidth': 1.0,
        'axes.labelweight': 'bold',
        'xtick.major.width': 1.0,
        'ytick.major.width': 1.0,
        'xtick.direction': 'out',
        'ytick.direction': 'out',
        'legend.frameon': False,
        'pdf.fonttype': 42,  # True Type fonts for PDFs
    }
)
```

### Diverging Colormap Centered on Zero

```python
# For data with meaningful zero point (e.g., log fold change)
from matplotlib.colors import TwoSlopeNorm

# Find data range
vmin, vmax = pivot_data.min().min(), pivot_data.max().max()
if not np.isfinite([vmin, vmax]).all() or not vmin < 0 < vmax:
    raise ValueError('TwoSlopeNorm with zero center needs finite values on both sides')
vcenter = 0

# Create norm
norm = TwoSlopeNorm(vmin=vmin, vcenter=vcenter, vmax=vmax)

# Plot
sns.heatmap(
    pivot_data,
    cmap='RdBu_r',
    norm=norm,
    annot=True,
    fmt='.2f'
)
```

## Large Datasets

### Downsampling Strategy

```python
# Equal allocation emphasizes rare groups and changes their apparent prevalence.
# Use full data for estimates/densities; label this as a display-only sample.
def smart_sample(df, target_size=10000, category_col=None, seed=7):
    if not isinstance(target_size, int) or target_size < 1:
        raise ValueError('target_size must be a positive integer')
    if len(df) <= target_size:
        return df.copy()
    if category_col is None:
        return df.sample(n=target_size, random_state=seed)
    groups = list(df.groupby(category_col, observed=True, dropna=False, sort=False))
    if target_size < len(groups):
        raise ValueError('Budget must allow at least one row per observed group')
    quota = target_size // len(groups)
    rng = np.random.default_rng(seed)
    return pd.concat([
        group.sample(n=min(len(group), quota), random_state=rng)
        for _, group in groups
    ])  # At most target_size; unused quota is deliberately not redistributed.

# Use sampled data for visualization
df_sampled = smart_sample(large_df, target_size=5000, category_col='category')

sns.scatterplot(data=df_sampled, x='x', y='y', hue='category', alpha=0.5)
```

### Hexbin for Dense Scatter Plots

```python
# For millions of points
fig, axes = plt.subplots(1, 2, figsize=(14, 6))

# Regular scatter (slow)
axes[0].scatter(df['x'], df['y'], alpha=0.1, s=1)
axes[0].set_title('Scatter (all points)')

# Hexbin (fast)
hb = axes[1].hexbin(df['x'], df['y'], gridsize=50, cmap='viridis', mincnt=1)
axes[1].set_title('Hexbin Aggregation')
plt.colorbar(hb, ax=axes[1], label='Count')

plt.tight_layout()
```

## Interactive Elements for Notebooks

These examples require an ipywidgets-enabled notebook frontend; they are illustrative and were not browser-tested. An empty filter should display a message rather than attempting a categorical plot.

### Adjustable Parameters

```python
from ipywidgets import interact, FloatSlider

@interact(bandwidth=FloatSlider(min=0.1, max=3.0, step=0.1, value=1.0))
def plot_kde(bandwidth):
    plt.figure(figsize=(10, 6))
    sns.kdeplot(data=df, x='value', hue='category',
                bw_adjust=bandwidth, fill=True)
    plt.title(f'KDE with bandwidth adjustment = {bandwidth}')
    plt.show()
```

### Dynamic Filtering

```python
from ipywidgets import interact, SelectMultiple

categories = df['category'].dropna().unique().tolist()
if not categories:
    raise ValueError('No nonmissing categories available')

@interact(selected=SelectMultiple(options=categories, value=[categories[0]]))
def filtered_plot(selected):
    filtered_df = df[df['category'].isin(selected)]
    if filtered_df.empty:
        print('Select at least one category with observations.')
        return

    fig, ax = plt.subplots(figsize=(10, 6))
    sns.violinplot(data=filtered_df, x='category', y='value', ax=ax)
    ax.set_title(f'Showing {len(selected)} categories')
    plt.show()
```


Sources: [Seaborn regression](https://seaborn.pydata.org/generated/seaborn.regplot.html),
[jointplot](https://seaborn.pydata.org/generated/seaborn.jointplot.html),
[clustermap](https://seaborn.pydata.org/generated/seaborn.clustermap.html),
[SciPy curve_fit](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.curve_fit.html),
[pandas groupby](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.groupby.html).
deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall