Data Viz Skill

This skill provides a standardized way to generate high-quality statistical charts for reports. It handles styling (using Sebrae-compatible colors), layout, and saving to files.

Capabilities

1. Bar Charts (plot_bar)

Best for comparing categories or counts. Supports vertical and horizontal orientation.

2. Pie Charts (plot_pie)

Best for showing composition (shares) of a whole. Limit to Top 5-7 categories for readability.

3. Grouped Bar Charts (plot_grouped_bar)

Best for comparing distributions across segments.

4. Evolution Line Charts (Survey Specific)

Best for comparing means of domains across points in time (e.g., Start vs End). Use plot_evolution_line.

5. Word Clouds

For qualitative text analysis visualization. Use in conjunction with survey-qual-analyzer frequencies.

6. Multivariate Analysis (principal_component_plotting, correlation_ellipse_plot, multivariate_normal_contours)

• PCA: Scree plots, loadings, and biplots.
• Correlation: Ellipse plots for correlation matrices (publication style).
• Probability: Bivariate normal distribution contours.

7. Performance Visualization (performance_curve_builder)

• Performance Curves: Cumulative results over time for any strategy.
• Drawdown: Visualizing risk periods and recovery.

Usage

import pandas as pd
# Import from skill scripts directory
from scripts.plotter import plot_bar, plot_pie, plot_grouped_bar
from scripts.evolution_plotter import plot_evolution_line
from scripts.visuals import *  # Additional visualization utilities
from scripts.advanced_plots import *  # Advanced chart types
from scripts.principal_component_plotting import plot_feature_importance, plot_biplot 
from scripts.correlation_ellipse_plot import plot_corr_ellipses
from scripts.multivariate_normal_contours import plot_contours
from scripts.performance_curve_builder import calculate_drawdown, extend_series_to_date

# 1. Simple Bar Chart (Top 10 Cities)
plot_bar(df, x_col="City", title="Respondents by City", filename="output/city_dist.png", orientation='h')

# 2. Evolution of Domains (Survey Pre vs Post)
# Expected columns: 'Cycle', 'Domain', 'Mean'
plot_evolution_line(df_evo, x="Cycle", y="Mean", hue="Domain", title="Evolution of Domains", filename="output/evolution.png")

Aesthetic Standards & Surveys

• Palette: Dark blue, cyan, and neutral grays for contrast.
• Labels: Always include sample size (n) if available.
• Premium: High DPI (300) and clean backgrounds for publication-ready reports.

Dependencies

Requires matplotlib, seaborn, and pandas.