Assignment 3: From data pipeline to report

Published

March 2, 2026

You built a data pipeline in class: download → clean → composite variable → aggregate → join GDP → scatter plot. Now turn that into a polished short report.

Step 1: Refine your pipeline and composite variable

Review and clean up your in-class code. Make sure it runs end-to-end from download to merged dataset.
Revisit your composite variable. Do all items point in the same direction? Should any be reversed before averaging? Justify your final selection of Q variables (3–6 items) in 2–3 sentences.
If you picked a single Q variable in class, now create a proper composite score (standardize each item to z-score, then average).

Step 2: Create exhibits

Think in terms of two variables: your composite score ($y$) and a GDP variable ($x$).

Graph 1: A carefully designed scatter plot showing the main relationship between income and your composite variable. Include informative title, axis labels, and country labels where useful.
Graph 2: Show heterogeneity by splitting countries into groups based on population size (e.g., small vs. large countries). This could be a colored scatter plot, faceted plot, or separate regression lines.
Table 1: Run a regression of your composite variable on log GDP per capita PPP. Report the coefficient, standard error, and interpretation. Use non-causal language — this is an association across countries.
- Stretch: Add population as a control. Does the coefficient change?

At each step, explain your choices and decisions (e.g., why you chose certain variables, why log GDP, what the population split means).

Step 3: Write the report

Use the exhbits. Check what AI does. Keep control but rely on it for formatting etc.

Step 4: Write the conclusion

Write a Conclusion paragraph (80–100 words) that summarizes: what you looked at, what you found, and what the limitations are. Avoid causal claims.

To submit

Report — maximum 2 pages in .pdf format, including exhibits (lastname_firstname_dawai_week03_report.pdf) (12p)
Code — your full pipeline from download to final graph and regression. Provide the file or a link. (2p)
Reflection — What advice would you give to a fellow data analysis student on using AI to create a report? Compare your different AI runs. What changes were decisive? How many iterations were useful? (lastname_firstname_dawai_week03_advice.txt) (6p)

Upload your report to the student folder called Reports at Moodle or similar service if applicable. This is for the next class — each group will read another group’s report and discuss its strengths and weaknesses.

AI use restrictions

For the report, use AI as your assistant — use it as input not as output. Don’t submit an AI-generated report without your own review and editing.
Do not use AI to generate the reflection. We want your personal examples and honest assessment, not AI-generated suggestions.

--- title: "Assignment 3: From data pipeline to report" date: "2026-03-02" --- You built a data pipeline in class: download → clean → composite variable → aggregate → join GDP → scatter plot. Now turn that into a polished short report. ### Step 1: Refine your pipeline and composite variable * Review and clean up your in-class code. Make sure it runs end-to-end from download to merged dataset. * Revisit your composite variable. Do all items point in the same direction? Should any be reversed before averaging? Justify your final selection of Q variables (3–6 items) in 2–3 sentences. * If you picked a single Q variable in class, now create a proper composite score (standardize each item to z-score, then average). ### Step 2: Create exhibits Think in terms of two variables: your composite score ($y$) and a GDP variable ($x$). * **Graph 1:** A carefully designed scatter plot showing the main relationship between income and your composite variable. Include informative title, axis labels, and country labels where useful. * **Graph 2:** Show heterogeneity by splitting countries into groups based on population size (e.g., small vs. large countries). This could be a colored scatter plot, faceted plot, or separate regression lines. * **Table 1:** Run a regression of your composite variable on log GDP per capita PPP. Report the coefficient, standard error, and interpretation. Use **non-causal** language — this is an association across countries. * **Stretch:** Add population as a control. Does the coefficient change? At each step, explain your choices and decisions (e.g., why you chose certain variables, why log GDP, what the population split means). ### Step 3: Write the report Use the exhbits. Check what AI does. Keep control but rely on it for formatting etc. ### Step 4: Write the conclusion Write a Conclusion paragraph (80–100 words) that summarizes: what you looked at, what you found, and what the limitations are. Avoid causal claims. ### To submit 1. **Report** — maximum 2 pages in .pdf format, including exhibits (`lastname_firstname_dawai_week03_report.pdf`) **(12p)** 2. **Code** — your full pipeline from download to final graph and regression. Provide the file or a link. **(2p)** 3. **Reflection** — What advice would you give to a fellow data analysis student on using AI to create a report? Compare your different AI runs. What changes were decisive? How many iterations were useful? (`lastname_firstname_dawai_week03_advice.txt`) **(6p)** Upload your report to the `student folder called Reports` at Moodle or similar service if applicable. This is for the next class — each group will read another group's report and discuss its strengths and weaknesses. ### AI use restrictions * For the report, use AI as your assistant — use it as *input* not as *output*. Don't submit an AI-generated report without your own review and editing. * Do not use AI to generate the reflection. We want your personal examples and honest assessment, not AI-generated suggestions.