Five Finger Exercises - Practice in Doing Data Wrangling / Engineering

[rufuspollock]

This is an issue for collecting initial data wrangling exercises for people to practice on.

Spending on Pharmaceuticals

NB: this exercise done see https://datahub.io/blog/pharmaceutical-drug-spending and https://github.com/datasets/awesome-data/issues/208

How much do we spend on pharmaceuticals - per country, over time? Also per capita and relative to other healthcare spending?

What are really expensive drugs?

https://data.oecd.org/healthres/pharmaceutical-spending.htm

Part I

• Download this data as CSV
• Turn into a data package
• Publish

But ... this does not have total.

To get that we'd need to multiple up e.g. spending per capita x population.

Part II

• Get a population time series dataset (preferably from OECD so we know same data) - or maybe we have a suitable one already in core datasets
• Store that as a data package

Part III

• Use these together to get total spending per country (and globally) - per year
• Get new data file and add that to the original spending data package

SELECT year, pop.population * pharma.spending_per_capita FROM ... JOIN ...

year, country, spendpercap

---

year, country, population

Document this process. Take screenshots. Imagine you were writing this up ...

Part III:

• Imagine we'd had first dataset and the population dataset already on DPR ...
• Could we have automated Part II

Building a "Data Package-ified" API for World Bank and OECD

For a world bank dataset:

• Script to turn the CSV version of the data into a Data Package
• Bonus: README

For an OECD dataset:

• Script to turn it into a Data Package