|
| 1 | +--- |
| 2 | +title: "Create a Data Cleaning System" |
| 3 | +format: pdf |
| 4 | +--- |
| 5 | + |
| 6 | +### Task |
| 7 | + |
| 8 | +During this assignment you are going to build a small system that cleans data and produces statistics |
| 9 | +automatically. |
| 10 | + |
| 11 | +Each group will have its own synthetic data set. |
| 12 | + |
| 13 | +The data set contains financial and economic data on companies. We are |
| 14 | +interested in publishing the average turnover from industrial activities, |
| 15 | +turnover from trade, and operating result, by 3-digit NACE code (the first three |
| 16 | +digits of the NACE variable). You can use e.g. `dplyr::summarise` to compute the |
| 17 | +results. |
| 18 | + |
| 19 | +You will build a set of scripts that cleans the raw data and estimates the desired quantities. |
| 20 | + |
| 21 | +In the afternoon each group will share its results by presenting their system. |
| 22 | +This may be a Powerpoint presentation but you can also just show the scripts and |
| 23 | +talk through it. |
| 24 | + |
| 25 | + |
| 26 | +### Tips |
| 27 | + |
| 28 | +- Look at the data. Make plots, discuss amongst each other what can be wrong with it, and how it might |
| 29 | + be solved. Don't try to solve everything at once. Solve one problem at the time. |
| 30 | +- Start by doing the estimates. They will be way off, but it is good to have a first result and see |
| 31 | +the effect of updates to the data cleaning process. |
| 32 | +- It is better to have simple code that runs than complicated code that doesn't. Start small, make sure it runs and then |
| 33 | +expand. |
| 34 | +- Use one script for each step in the statistical value chain. Each script reads an input, does something to the data, and writes an output. |
| 35 | +- Define rules to check the quality of the data. Also here: work iteratively. Start with a few rules |
| 36 | + and adapt the ruleset iteratively. |
| 37 | +- Iterate often, view the data and the results often. |
| 38 | +- Make plots of the data: are there outliers? Also think about ratios between variables. |
| 39 | +- Impute the missing values. Try a few models. |
| 40 | + |
| 41 | + |
| 42 | + |
| 43 | +### The data |
| 44 | + |
| 45 | + |
| 46 | +The financial variabeles have to satisfy the following balance restrictions. You can think of extra |
| 47 | +restrictions as you deem fit. |
| 48 | +\begin{center} |
| 49 | +\includegraphics[width=\textwidth]{balances.pdf} |
| 50 | +\end{center} |
| 51 | + |
| 52 | + |
| 53 | + |
0 commit comments