---
title: "Metadata Vocabularies for Input–Output Analysis"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Metadata Vocabularies for Input–Output Analysis}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

```{r setup}
library(iotables)
```

## Import and Normalisation Workflow

The five Eurostat vocabularies —\
`ind_ava`, `ind_use`, `prd_ava`, `prd_use`, and `cpa2_1` — were imported directly from the official Eurostat metadata registry (<https://dd.eionet.europa.eu/vocabulary/eurostat/>).\
Each dataset mirrors the structure of its corresponding SDMX codelist and preserves Eurostat’s identifiers and validity information.

### Data Sources

| Vocabulary | Description | Source URL |
|------------------------|------------------------|------------------------|
| `ind_ava` | Industries, adjustments and value added (rows for industry × industry SIOTs) | <https://dd.eionet.europa.eu/vocabulary/eurostat/ind_ava/> |
| `ind_use` | Industry uses (columns for industry × industry SIOTs) | <https://dd.eionet.europa.eu/vocabulary/eurostat/ind_use/> |
| `prd_ava` | Products, adjustments and value added (rows for product × product SIOTs) | <https://dd.eionet.europa.eu/vocabulary/eurostat/prd_ava/> |
| `prd_use` | Product uses (columns for product × product SIOTs) | <https://dd.eionet.europa.eu/vocabulary/eurostat/prd_use/> |
| `cpa2_1` | Statistical Classification of Products by Activity (CPA 2.1) | <https://dd.eionet.europa.eu/vocabulary/eurostat/cpa2_1/> |

### Import Steps

1.  **Raw Download**\
    Each vocabulary was retrieved as an Excel export from EIONET’s vocabulary registry.

2.  **Column Standardisation**\
    Columns were renamed to a unified schema: `id, label, status, status_modified, notation, group, quadrant, numeric_order, iotables_label, block, uri`

3.  **Quadrant and Block Assignment**\
    Each item was assigned a `quadrant` and a semantic `block` consistent across vocabularies:

    -   `10` = intermediate (Quadrant 1)

    -   `20` = primary_inputs (Quadrant 3)

    -   `30` = final_use (Quadrant 2)

    -   `50` = extension / diagnostic\
        Control totals such as *“Total supply at basic prices”* were retained as `block = "control_total"`.

4.  **Ordinal Ordering**\
    `numeric_order` was reindexed within each quadrant with consistent gaps (10, 20, …) to ensure reproducible ordering for matrix construction.

5.  **URI Generation**\
    Each code was linked to its SKOS concept using:

```{r uri, eval=FALSE}
df$uri <- sprintf(
  "https://dd.eionet.europa.eu/vocabularyconcept/eurostat/%s/%s",
  vocabulary_id,
  df$notation
)
```

6.  **Validation**\
    Each table was checked for:

    -   missing or duplicate IDs

    -   monotone numeric order

    -   alignment of quadrant ↔ block semantics

7.  **Storage and Naming**

The cleaned tibbles were stored as exported data objects:

```         
data/ind_ava.rda data/ind_use.rda data/prd_ava.rda data/prd_use.rda 
```

Each dataset can be loaded directly with `data(<name>)`.

### Adjustments to Vocabularies

Although the four Eurostat vocabularies (`ind_ava`, `ind_use`, `prd_ava`, `prd_use`) were imported directly from the official Eurostat metadata registry, some modifications were necessary to ensure compatibility with the actual Eurostat input–output datasets. The main data sources, in particular `naio_10_cp1750` and `naio_10_cp1700`, occasionally include variables that are not coded according to the published and standardised vocabularies. While these inconsistencies are usually clear to a manual user, they can create ambiguity in a reproducible workflow where automated matching is required.

For example, the *product × product* SIOTs for the Slovak Republic contain a more detailed industry breakdown than that defined in prd_ava and prd_use. To maintain alignment across datasets, all 0-, 1-, and 2-digit codes from the `cpa2_1` vocabulary were imputed into the four vocabularies. Each entry includes a validity flag in the status column, indicating whether the code is valid in the official Eurostat vocabulary or was adopted from observed but non-standard codes in the data. This approach preserves reproducibility while ensuring complete coverage of all codes encountered in current Eurostat data releases.

### Versioning

All four vocabularies correspond to the 2025 Eurostat CPA 2.1 / ESA 2010 edition.