Dealing with the change of administrative divisions over time with

class: center, middle, inverse, title-slide

# Dealing with the change of administrative divisions over time with 
### Kim Antunez antuki13
### useR! 2019 - Toulouse, France 11 July 2019

---

## Administrative divisions...

.left-column[
<img src="img/millefeuille2.png" height="60px"></img>
<img src="img/millefeuille1.png" height="400px">
]
.right-column[
<center>The French territorial « mille-feuille »</center>
<center><img src="img/carte_millefeuille_v2.png"></center>
]

---
## ... change over time
<center><img src="img/map_fusions.png" height="500px"></center>

---
.pull-left[
### See modifications over the years
]

.pull-right[
 <center><img src="img/trajectoires.png" height="100px"></center>
]

All functions and parameters are translated from French for this presentation

```r
municipality_evolution_graph(code = "76108", year = 2014) # uses {visNetwork}
```

```r
evol <- municipalities_evolutions(begin_date="01-01-2011", end_date="01-01-2014")
```

```r
grep("(76095)|(76108)", evol$mergers, value = TRUE)
```

```
## 2012-01-01: Bois-Guillaume-Bihorel (76108) is a merger of Bihorel (76095),
Bois-Guillaume (76108).
```

```r
grep("(76095)|(76108)", evol$divisions, value = TRUE)
```

```
## 2014-01-01: Bois-Guillaume (76108) divided into Bois-Guillaume (76108), Bihorel
## (76095).
```

---

.pull-left[
### Guess the year of a database
]

.pull-right[
<center><img src="img/tableau_year.png" height="100px"></center>
]

```r
head(db, 2)
```

```
##    code      men   women
## 1 01001 385.0000 380.000
## 2 01002 142.6132 113.107
```

```r
OGC_guess(db$code) # guesses the year of the Official Geographic Code (OGC)
```

```
## [1] "2016"
```

```r
codes <- c("99086", db$code[-1]) # creates a vector of ID of municipalities
merge_OGC(codes = codes, OGC = 2016)$not_in_db # ID not in the db
```

```
## 01001
```

```r
merge_OGC(codes = codes, OGC = 2016)$not_in_OGC # ID not in the OGC
```

```
## 99086
```

---
.pull-left[
### Change the year of a database

* **qualitative variable** [character] 
 See `change_OGC_typology`

+ *divisions * easy: copy the lines
 + *mergers * several hypotheses: assign the class that contains the most population, define an absorbent or absorbed class...
]

.pull-right[
 <center><img src="img/tableau_year2.png" height="100px"></center>

* **quantitative variable** [numeric]
 See `change_OGC_numeric`

+ *mergers * easy: sum the lines 
 + *divisions * divide lines proportionally to population

]

```r
nrow(db)
```

```
## [1] 35887
```

```r
db_2019 <- db %>% # changes the year of a numeric variable (from 2016 to 2019)
 change_OGC_numeric(2016:2019)
```

```r
str(db_2019)
```

```
## 'data.frame':	34972 obs. of  3 variables:
##  $ code : chr  "01001" "01002" "01004" "01005" ...
##  $ men  : num  385 142.6 6778.5 819 54.5 ...
##  $ women: num  380 113.1 7246.5 783.9 49.5 ...
```
---
.pull-left[
### Aggregate a database
]

.pull-right[
<center><img src="img/agreger.png" height="100px"></center>
]

```r
db_dep <- db_2019 %>% # aggregates the database (municipalities -> départements)
 aggregate_OGC(OGC = 2019, administrative_division = "DEP")
```

```r
str(db_dep)
```

```
## 'data.frame':	100 obs. of  4 variables:
##  $ DEP   : chr  "01" "02" "03" "04" ...
##  $ LIBGEO: chr  "Ain" "Aisne" "Allier" "Alpes-de-Haute-Provence" ...
##  $ men   : num  309198 262693 163759 78600 68309 ...
##  $ women : num  316839 276817 179006 83124 71588 ...
```

```r
DEP_sf <- left_join(DEP_sf,db_dep,by=c("DEP"="DEP")) %>%
 mutate(prop=100*men/(men+women))
plot(DEP_sf %>% select(prop))
```

---

## Future improvements?

* **Create new functions**
    +  add your own geographical levels
    +  add your own distribution keys
    
* **Expand to other countries**
    +  in Europe : Nomenclature of Territorial Units for Statistics (NUTS)
    +  identify common functionnalities VS local (French!) specificities

* **Reach non R-users**
    +  Shiny Apps
    +  API

---

class: center, middle

<a href="http://twitter.com/antuki13" class="social">@antuki13</a>
<a href="http://github.com/antuki" class="social">antuki</a>
<a href="http://antuki.github.io" class="social">antuki.github.io</a>

**Packages on github: [antuki/COGugaison](https://github.com/antuki/COGugaison) and [antuki/CARTElette](https://github.com/antuki/CARTElette).**

Slides created with R package [**xaringan**](https://github.com/yihui/xaringan) with the [R-Ladies theme](https://alison.rbind.io/post/r-ladies-slides/).

Also with [remark.js](https://remarkjs.com), [knitr](http://yihui.name/knitr), and [R Markdown](https://rmarkdown.rstudio.com).