Panel Plots

Panel structure plots help you understand the characteristics of your panel data: entity-specific effects, time trends, variance decomposition, and data balance. PanelBox provides 4 interactive charts tailored for panel data analysis.

Quick Start

from panelbox.visualization import create_panel_charts

# Create all 4 panel charts from model results
charts = create_panel_charts(results, theme="professional", include_html=False)

# Access individual charts
charts["entity_effects"].to_html()
charts["structure"].save_image("balance.png")

You can also select specific chart types:

charts = create_panel_charts(
    results,
    chart_types=["entity_effects", "between_within"],
    theme="academic",
    include_html=False,
)

Entity Effects Plot

Registry name: panel_entity_effects

Horizontal bar chart of entity-specific fixed or random effects with optional confidence intervals. Helps identify which entities deviate most from the overall mean.

from panelbox.visualization import create_entity_effects_plot

chart = create_entity_effects_plot(
    panel_results,
    theme="professional",
    sort_by="magnitude",      # 'magnitude', 'alphabetical', 'significance'
    max_entities=20,          # Limit display for large panels
    show_confidence=True,
)
chart.figure.show()

Using the factory directly:

from panelbox.visualization import ChartFactory

chart = ChartFactory.create(
    "panel_entity_effects",
    data={
        "entity_id": ["Firm A", "Firm B", "Firm C", ...],
        "effect": [0.23, -0.15, 0.08, ...],
        "std_error": [0.05, 0.04, 0.06, ...],   # Optional
    },
    theme="academic",
)

Interpretation:

• Large positive effects: entities with higher-than-average outcomes
• Large negative effects: entities with lower-than-average outcomes
• Wide confidence intervals: imprecise estimates, possibly few observations for that entity
• Clustering of effects: groups of similar entities

Time Effects Plot

Registry name: panel_time_effects

Line chart of time fixed effects with confidence bands and significance highlighting. Shows how the outcome variable trends over time after controlling for entity effects and covariates.

from panelbox.visualization import create_time_effects_plot

chart = create_time_effects_plot(
    panel_results,
    theme="professional",
    show_confidence=True,
    highlight_significant=True,
)

Using the factory directly:

chart = ChartFactory.create(
    "panel_time_effects",
    data={
        "time": [2000, 2001, 2002, 2003, ...],
        "effect": [0.0, 0.12, 0.25, 0.18, ...],
        "std_error": [0.03, 0.03, 0.04, 0.03, ...],   # Optional
    },
    theme="academic",
)

Interpretation:

• Upward trend: positive time effects, possibly reflecting macro trends
• Sudden jumps: structural breaks or policy changes
• Significant periods (starred): time effects statistically different from zero
• Confidence band width: precision of time effect estimates

Between-Within Plot

Registry name: panel_between_within

Variance decomposition showing how much of each variable's variation is between entities versus within entities over time.

from panelbox.visualization import create_between_within_plot

chart = create_between_within_plot(
    panel_data,
    variables=["wage", "education", "experience"],
    theme="academic",
    style="stacked",          # 'stacked', 'side_by_side', or 'scatter'
    show_percentages=True,
)

Three visualization styles are available:

Stacked BarsSide-by-Side BarsScatter Plot

chart = create_between_within_plot(data, style="stacked")

Stacked bar chart showing between and within variance as proportions of total variance.

chart = create_between_within_plot(data, style="side_by_side")

Grouped bar chart for direct comparison of absolute variance magnitudes.

chart = create_between_within_plot(data, style="scatter")

Scatter plot of between vs within variance, with a 45-degree reference line.

Using the factory directly:

chart = ChartFactory.create(
    "panel_between_within",
    data={
        "variables": ["wage", "education", "experience"],
        "between_var": [2.5, 1.8, 3.2],
        "within_var": [1.2, 0.4, 2.1],
    },
    theme="professional",
    config={"chart_type": "stacked"},
)

Interpretation:

• High between-variation: variable differs mainly across entities (cross-sectional variation dominates)
• High within-variation: variable changes mainly over time within entities (temporal variation dominates)
• Implications for estimation:
  • Fixed Effects exploits within-variation only
  • Random Effects uses both, but requires exogeneity
  • Between Estimator uses between-variation only

Panel Structure Plot

Registry name: panel_structure

Heatmap showing data presence/absence across entities and time periods. Essential for diagnosing unbalanced panels and understanding missing data patterns.

from panelbox.visualization import create_panel_structure_plot

chart = create_panel_structure_plot(
    panel_data,
    theme="presentation",
    show_statistics=True,
    highlight_complete=True,
)

Using the factory directly:

chart = ChartFactory.create(
    "panel_structure",
    data={
        "entities": ["Firm A", "Firm B", "Firm C"],
        "time_periods": [2000, 2001, 2002, 2003],
        "presence_matrix": [
            [1, 1, 1, 1],   # Firm A: complete
            [1, 1, 0, 1],   # Firm B: missing 2002
            [0, 1, 1, 1],   # Firm C: missing 2000
        ],
    },
    theme="professional",
)

Interpretation:

• All green: balanced panel (all entities observed in all periods)
• Red cells: missing observations, investigate the pattern
• Systematic gaps: attrition, late entry, or structural missing data
• Random gaps: may be ignorable under MAR assumption

Data Transformers

The PanelDataTransformer extracts panel-specific data from model results:

from panelbox.visualization.transformers.panel import PanelDataTransformer

# Extract entity effects from model results
entity_data = PanelDataTransformer.extract_entity_effects(panel_results)

# Extract time effects
time_data = PanelDataTransformer.extract_time_effects(panel_results)

# Calculate between-within variance
bw_data = PanelDataTransformer.calculate_between_within(
    panel_data, variables=["wage", "education"]
)

# Analyze panel structure
structure = PanelDataTransformer.analyze_panel_structure(panel_data)

Complete Example

Full panel structure analysis workflow:

import panelbox as pb
from panelbox.visualization import (
    create_panel_charts,
    create_between_within_plot,
    export_charts,
)

# Load data and estimate model
data = pb.load_dataset("wagepan")
model = pb.FixedEffects(data=data, formula="lwage ~ hours + exper + EntityEffects")
results = model.fit()

# Generate all panel charts
charts = create_panel_charts(results, theme="professional", include_html=False)

# Create additional variance decomposition
bw_chart = create_between_within_plot(
    data,
    variables=["lwage", "hours", "exper"],
    theme="academic",
    style="stacked",
    show_percentages=True,
)

# Export everything
export_charts(charts, output_dir="./panel_analysis", format="png", width=900, height=600)
bw_chart.save_image("panel_analysis/variance_decomposition.png")

Comparison with Other Software

Chart	PanelBox	Stata	R
Entity effects	panel_entity_effects	Manual after xtreg	Custom ggplot
Time effects	panel_time_effects	Manual after xtreg	Custom ggplot
Variance decomposition	panel_between_within	xtsum (table only)	Custom calculation
Panel balance	panel_structure	xtset, xtdescribe	plm::pdim(), plm::pbalanced()

See Also

• Diagnostics Plots -- Residual diagnostics for panel models
• Comparison Plots -- Compare FE, RE, and Pooled OLS
• Themes & Customization -- Style your panel charts
• Test Plots -- Hausman test and other panel diagnostics