data-analyst

name	data-analyst
description	SQL queries, spreadsheet analysis, charts, and statistical methods. Use when the user has data to analyze, needs visualizations, or wants insights from CSV/JSON/Excel files.
version	1.0.0
license	MIT
compatibility	Python 3.8+ with pandas, matplotlib, seaborn, scipy
metadata	{"author":"hermeshub","hermes":{"tags":["data-analysis","sql","charts","statistics","visualization"],"category":"data","requires_tools":["terminal"]}}

Data Analyst

End-to-end data analysis with visualization and reporting.

When to Use

• User provides a dataset (CSV, JSON, Excel, SQLite)
• User asks for data exploration, trends, or patterns
• User needs charts, graphs, or visualizations
• User wants statistical analysis or hypothesis testing
• User asks for a summary report from data

Procedure

1. Load and inspect the data (shape, dtypes, nulls, head)
2. Clean: handle missing values, fix types, remove duplicates
3. Explore: distributions, correlations, outliers
4. Analyze: answer the specific question or find patterns
5. Visualize: create appropriate charts
6. Report: structured findings with actionable insights

Analysis Toolkit

Quick Stats

import pandas as pd
df = pd.read_csv("data.csv")
print(df.describe())
print(df.info())
print(df.isnull().sum())

Visualization

import matplotlib.pyplot as plt
import seaborn as sns
# Distribution
sns.histplot(df['column'], kde=True)
# Correlation
sns.heatmap(df.corr(), annot=True)
# Time series
df.plot(x='date', y='value', figsize=(12,6))
plt.savefig('chart.png', dpi=150, bbox_inches='tight')

Statistical Tests

from scipy import stats
# T-test
t_stat, p_val = stats.ttest_ind(group_a, group_b)
# Correlation
r, p = stats.pearsonr(x, y)

Output Format

• Always start with a data summary (rows, columns, types)
• Show key statistics before diving into analysis
• Every chart must have title, axis labels, and legend
• End with actionable recommendations

Pitfalls

• Always check for null values before calculations
• Verify data types (strings disguised as numbers)
• Watch for survivorship bias in time series
• State sample sizes and confidence intervals
• Don't confuse correlation with causation

Verification

• Row counts match expected after cleaning
• Charts render correctly and save to disk
• Statistical results include p-values and effect sizes