5-Day Workshop: Data Analysis Using AI Prompts
Comprehensive Training Manual
1 🧠 Workshop Overview
This 5-day workshop focuses entirely on learning how to use AI prompts for data analysis, regardless of whether the analysis is done in R, Python, Stata, SPSS, Excel, or Tableau.
The workshop emphasizes:
- Prompt engineering for data
- Analysis roadmap prompts
- EDA prompts
- Survey & NHANES prompts
- Modeling prompts
- Mapping prompts
- Storytelling prompts
- Debugging prompts
- Vibe-coding prompts
- Case-based scenario prompts
2 Day 1 — Foundations of AI Prompting for Data Analysis
2.1 1.1 What is a Data Prompt?
A data prompt guides AI to produce relevant, accurate, and structured analysis.
2.2 1.2 CRISP Prompting Method
C — Context
R — Role
I — Input
S — Steps
P — Preferred Output
2.2.1 Template
You are a data analysis expert.
Context: [dataset]
Task: [goal]
Steps: [analysis steps]
Output: [tables, plots, bullet points]2.3 1.3 Core Prompts for Day 1
2.3.1 🔍 Dataset Understanding
Provide structure, variable types, missingness, and 10 insights from this dataset.2.3.2 📊 EDA Roadmap
Create a complete EDA plan with summaries, grouped analysis, and recommended plots.2.3.3 📁 Data Dictionary
Generate a data dictionary for all variables, including definitions and plausible ranges.3 Day 2 — Survey, Health, Demographic & NHANES Prompting
3.1 2.1 Survey Data Prompts
Summarize this survey dataset: sampling design, weights, PSUs, strata, key demographics.3.2 2.2 NHANES Prompts
3.2.1 Cleaning & Merging
Clean NHANES data: demographics, labs, exams. Create BMI category, BP groups, race labels.3.2.2 Risk Index
Create a composite risk index using hypertension, obesity, diabetes, smoking, inactivity.3.2.3 Survey-Weighted Model
Fit a weighted logistic regression predicting hypertension and interpret results simply.4 Day 3 — Modeling Prompts
4.1 3.1 Universal Model Selection
Choose the best model for this outcome. Compare regression, logistic, random forest, boosting, and explain decision.4.2 3.2 Regression Interpretation
Fit a regression model and explain coefficients in simple English with policy insights.4.3 3.3 Classification
Build a classification model. Provide confusion matrix, feature importance, and interpretation.4.4 3.4 Time-Series
Analyze long-term trends, produce forecasts, and describe patterns.4.5 3.5 Panel Data
Fit FE and RE models, run Hausman test, and interpret results for policymakers.5 Day 4 — Maps, Storytelling, Visualization Prompts
5.1 4.1 Mapping Prompts
Join spatial data with indicators and create a choropleth map. Explain spatial variation.5.2 4.2 Visualization Enhancement
Improve this plot: labels, theme, color scheme, clarity, interpretability.5.3 4.3 Storytelling Prompts
5.3.1 Narrative Writing
Write a three-paragraph narrative explaining the findings in simple English.5.3.2 Executive Summary
Write a one-page executive summary with key findings, implications, and recommendations.5.3.3 LinkedIn Post
Convert this analysis into a concise, engaging LinkedIn post.6 Day 5 — Debugging, Vibe Coding, Scenario Practice
6.1 5.1 Debugging
Find the error in this analysis pipeline, explain the reason, and provide corrected code/output.6.2 5.2 Vibe Coding
Rewrite my analysis code to be cleaner, more readable, modular, and commented.6.3 5.3 End-to-End Project Prompt
Produce an entire analysis workflow:
cleaning → exploration → modeling → mapping → insights → narrative.7 Deliverables
- Prompt libraries
- Case studies
- Cheat sheets
- AI-generated reports
- Executive summaries
- Mapping prompts
- Modeling prompts
8 Appendix — Master Prompt Library (100+ Prompts)
8.1 Dataset Prompts
Explain dataset in 10 bullet points.
Find anomalies and inconsistencies.
Generate 10 research questions from this dataset.8.2 Survey Prompts
Produce weighted estimates and explain design effects.8.3 Modeling Prompts
Run logistic regression and interpret ORs simply.
Explain model uncertainty.8.4 Storytelling Prompts
Turn results into 5 policy recommendations.8.5 Debug Prompts
Why is this model output unrealistic? Diagnose and fix.8.6 Vibe Coding Prompts
Modernize my code using tidyverse/pythonic structure.