應用計量經濟學導論

class: center, middle, inverse, title-slide

.title[
# 應用計量經濟學導論
]
.subtitle[
## 實習課
]
.author[
### <div style="margin-top:-18px; line-height:1.2;">
<p><span style="font-weight:700;">王俊欽</span><br />
<span style="font-weight:700;"><a href="mailto:d14627001@ntu.edu.tw" class="email">d14627001@ntu.edu.tw</a></span><br />
<span style="font-weight:700;">2026/09/21</span><br />
 </p>
</div>
]

---

layout: true
background-image: url(data:image/png;base64,#https://www.agec.ntu.edu.tw/uploads/asset/data/675f864c74ccc38a768f8254/%E6%9C%AA%E4%BE%86%E5%B1%95%E6%9C%9B%E5%9C%96%E7%89%87.png)
background-position: 98% 2%
background-size: 5%

---
## Recall pipeline

- ` %>% ` 是 tidy 家族中常用且使程式碼更易閱讀的運算子

- 使用 pipeline 建立工作流程（workflow）有效建立操作邏輯

``` r
library(dplyr) 
```

- 預設是將左方 R object 帶入右方指令的第一個參數

.pull-left[
f(x)

``` r
sum(c(1:5))
```

```
## [1] 15
```
]

.pull-right[
x %>% f()

``` r
c(1:5) %>% sum()
```

```
## [1] 15
```
]

---
## Recall pipeline
**penguins** 資料集中包含企鵝品種、居住島、喙長寬與體重等等變數。

``` r
head(penguins)
```

```
##   species    island bill_len bill_dep flipper_len body_mass    sex year
## 1  Adelie Torgersen     39.1     18.7         181      3750   male 2007
## 2  Adelie Torgersen     39.5     17.4         186      3800 female 2007
## 3  Adelie Torgersen     40.3     18.0         195      3250 female 2007
## 4  Adelie Torgersen       NA       NA          NA        NA   <NA> 2007
## 5  Adelie Torgersen     36.7     19.3         193      3450 female 2007
## 6  Adelie Torgersen     39.3     20.6         190      3650   male 2007
```
---
## Recall pipeline
使用 dplyr 指令可以讓工作流程更清晰，例如：統計不同品種、性別間的企鵝嘴喙的長寬比最小值是多少。
.panelset[
.panel[.panel-name[w/o dplyr]

``` r
summarize(select(group_by(mutate(drop_na(
          penguins), bill_ratio = bill_len/ bill_dep), species, sex), species, sex, bill_ratio), min_bill_ratio = min(bill_ratio), .groups = "drop")
```

``` r
penguins %>%
  drop_na() %>%
  mutate(bill_ratio = bill_len/ bill_dep) %>% 
  group_by(species, sex) %>%
  select(species, sex, bill_ratio) %>% 
  summarize(min_bill_ratio = min(bill_ratio), .groups = 'drop')
```

```
## # A tibble: 6 × 3
##   species   sex    min_bill_ratio
##   <fct>     <fct>           <dbl>
## 1 Adelie    female           1.76
## 2 Adelie    male             1.64
## 3 Chinstrap female           2.35
## 4 Chinstrap male             2.48
## 5 Gentoo    female           2.84
## 6 Gentoo    male             2.57
```
]
]
---
## Recall pipeline
<img src="data:image/png;base64,#./Pipe_baking_magrittr_backAssign.gif" width="55%" style="display: block; margin: auto;" />
<p style="text-align: center;">圖片來源：<a href="https://github.com/arthurwelle/VIS/blob/master/Pipe_Cake/Pipe_baking_magrittr_backAssign.gif">Arthur Welle</a></p>

---
## Merge
通常在整理資料的過程中，需要透過結合各個不同類型或是時間段的資料並加以分析，這時就必須將多個資料進行合併。實務上，資料合併大致可分為：

- 垂直合併：通常在合併相同但**時間不同**的資料集，或是合併**不同公司或單位**的財務報表或股價資料，這些資料集通常都會有相同的欄位（column）名稱與數量，整體資料寬度不會改變，例如：將一月至十二月的天氣資料垂直合併。

- 水平合併：通常在合併有關連的資料集，這類資料通常主題或調查製作單位不一樣，兩個資料集的行數或欄位皆可以不盡相同但是必須**具有一個或多個索引鍵（key value）**使兩個資料集有依據的水平合併起來，例如：將國家總經資料與社經資料依照國家名稱水平合併。

---
## Vertically Merge
將具有相同欄位數量與名稱（非必要）的資料以上下疊加的方式垂直合併。
.panelset[
.panel[.panel-name[rbind]
.pull-left[

``` r
d1 <- tribble(
  ~id, ~name, ~score,
  1, "Chloe", 95,
  2, "Leo"  , 60,
  3, "Sandy", 82
)
d2 <- tibble(
  id = c(4:6),
  name = c("Tim", "Irene", "Cindy"),
  score = id * 10 + 15
)
```
]
.pull-right[

``` r
rbind(d1, d2)
```

```
## # A tibble: 6 × 3
##      id name  score
##   <dbl> <chr> <dbl>
## 1     1 Chloe    95
## 2     2 Leo      60
## 3     3 Sandy    82
## 4     4 Tim      55
## 5     5 Irene    65
## 6     6 Cindy    75
```
]
]
.panel[.panel-name[difference column]
當資料集中存在不同變數名稱時，未合併到的部分將為 NA。
.pull-left[

``` r
d3 <- tribble(
  ~id, ~name, ~math,
  1, "Chloe", 95,
  2, "Leo"  , 60,
  3, "Sandy", 82
)
d4 <- tibble(
  id = c(4:6),
  name = c("Tim", "Irene", "Cindy"),
  english = id * 10 + 15
)
```
]
.pull-right[

``` r
d3 %>% bind_rows(d4)
```

```
## # A tibble: 6 × 4
##      id name   math english
##   <dbl> <chr> <dbl>   <dbl>
## 1     1 Chloe    95      NA
## 2     2 Leo      60      NA
## 3     3 Sandy    82      NA
## 4     4 Tim      NA      55
## 5     5 Irene    NA      65
## 6     6 Cindy    NA      75
```
]
]
]

---
## Vertically Merge
如果兩個資料集的變數名稱不同時，需要先將變數名稱修改一致，再進行垂直合併。
.panelset[
.panel[.panel-name[data]
.pull-left[

``` r
d5 <- tribble(
  ~order, ~name, ~math,
  1, "Chloe", 95,
  2, "Leo"  , 60,
  3, "Sandy", 82
)
```

```
## # A tibble: 3 × 3
##   order name   math
##   <dbl> <chr> <dbl>
## 1     1 Chloe    95
## 2     2 Leo      60
## 3     3 Sandy    82
```
]
.pull-right[

``` r
d6 <- tibble(
  id = c(4:6),
  name = c("Tim", "Irene", "Cindy"),
  english = id * 10 + 15
)
```

```
## # A tibble: 3 × 3
##      id name  english
##   <int> <chr>   <dbl>
## 1     4 Tim        55
## 2     5 Irene      65
## 3     6 Cindy      75
```
]
]
.panel[.panel-name[rename]

``` r
d5 %>% rename(id = order) %>% bind_rows(d6)
```

```
## # A tibble: 6 × 4
##      id name   math english
##   <dbl> <chr> <dbl>   <dbl>
## 1     1 Chloe    95      NA
## 2     2 Leo      60      NA
## 3     3 Sandy    82      NA
## 4     4 Tim      NA      55
## 5     5 Irene    NA      65
## 6     6 Cindy    NA      75
```
]
]

---
## Horizontally Merge
通常在整理資料的過程中，需要結合各個不同類型的資料並加以分析，這時需要依據**索引鍵（key value）**將不同資料水平合併，tidyverse 中有幾種 join 方式大致可分為：
.left-column[

- Full Join

- Left Join

- Right Join

- Inner Join

- Semi Join
]
.right-column[
<img src="data:image/png;base64,#./tidyversrejoin.png" width="75%" style="display: block; margin: auto;" />
]

---
## Horizontally Merge
有兩個資料集，資料集 people 中包含個人的 id, name 兩個變數；資料集 school 中包含個人 id, school 兩個變數，我們可以藉由索引鍵 **id** 來水平合併兩個資料集。

``` r
people <- tibble(
  id = c(1, 2, 3, 4, 6),
  name = c("Tim", "Irene", "Cindy", "Chloe", "Sandy")
  
)
school <- tribble(
  ~id, ~school,
  3, "NTU",
  4, "NTNU",
  5, "NCCU",
  7, "NTU"
)
```

---
## Horizontally Merge
依據一或多個索引鍵（key value）對兩個或以上的資料集進行水平合併， `by` argument 可以提供實際要使用的索引鍵，沒設定的話會自動依據被合併的資料集做選擇。
.panelset[
.panel[.panel-name[Full]

``` r
people %>% full_join(school, by = "id")
```

```
## # A tibble: 7 × 3
##      id name  school
##   <dbl> <chr> <chr> 
## 1     1 Tim   <NA>  
## 2     2 Irene <NA>  
## 3     3 Cindy NTU   
## 4     4 Chloe NTNU  
## 5     6 Sandy <NA>  
## 6     5 <NA>  NCCU  
## 7     7 <NA>  NTU
```
]
.panel[.panel-name[Left]
以左表 people 為主表，將右表 school 向左水平合併。

``` r
people %>% left_join(school, by = "id")
```

```
## # A tibble: 5 × 3
##      id name  school
##   <dbl> <chr> <chr> 
## 1     1 Tim   <NA>  
## 2     2 Irene <NA>  
## 3     3 Cindy NTU   
## 4     4 Chloe NTNU  
## 5     6 Sandy <NA>
```
]
.panel[.panel-name[Right]
以右表 school 為主表，將左表 people 向右水平合併。

``` r
people %>% right_join(school, by = "id")
```

```
## # A tibble: 4 × 3
##      id name  school
##   <dbl> <chr> <chr> 
## 1     3 Cindy NTU   
## 2     4 Chloe NTNU  
## 3     5 <NA>  NCCU  
## 4     7 <NA>  NTU
```
]
.panel[.panel-name[Inner]
沒設定 by 什麼索引鍵的話，則自動依據被合併的資料集做選擇。

``` r
people %>% inner_join(school)
```

```
## Joining with `by = join_by(id)`
```

```
## # A tibble: 2 × 3
##      id name  school
##   <dbl> <chr> <chr> 
## 1     3 Cindy NTU   
## 2     4 Chloe NTNU
```
]
]

---
## Horizontally Merge

`semi_join()` 與 `inner_join()` 差別在於 `semi_join()` 不會保留右表資料。
.pull-left[
<img src="data:image/png;base64,#./semi-join.gif" width="100%" style="display: block; margin: auto;" />
]
.pull-right[
<img src="data:image/png;base64,#./inner-join.gif" width="100%" style="display: block; margin: auto;" />
]
<p style="text-align: center;">圖片來源：<a href="https://github.com/gadenbuie/tidyexplain">Garrick Aden-Buie tidyexplain</a></p>
---
## Horizontally Merge

`semi_join(x, y)` 是當遇到一種情況：舉行交換學生行前說明會之前，想從學生名單中挑出**有錄取的交換學生**寄通知信。此時我只需要用錄取名單中的 `id` 來篩選左表，而不需要學校、國家、學期等欄位資訊。

``` r
people <- tibble(
  id    = c(1, 2, 3, 4, 6),
  name  = c("Tim", "Irene", "Cindy", "Chloe", "Sandy"),
  score = c(88, 76, 95, 60, 82)
)

exchange <- tribble(      
  ~id, ~school, ~country,  ~semester,
  1,   "NTU",   "Japan",   "2025 Fall",
  3,   "NTNU",  "Germany", "2025 Fall",
  6,   "NCCU",  "American",   "2026 Spring",
  7,   "NTU",   "American",  "2026 Spring"
)
```

---
## Horizontally Merge
.panelset[
.panel[.panel-name[Semi]
從左表 people 中篩選出 id 也在右表 exchange 的資料，篩選完後只保留左表資料。

``` r
people %>% semi_join(exchange, by = "id")
```

```
## # A tibble: 3 × 3
##      id name  score
##   <dbl> <chr> <dbl>
## 1     1 Tim      88
## 2     3 Cindy    95
## 3     6 Sandy    82
```
]
.panel[.panel-name[Anti]
從左表 people 中篩選出 id 不在右表 exchange 的資料，篩選完後只保留左表資料。

``` r
people %>% anti_join(exchange, by = "id")
```

```
## # A tibble: 2 × 3
##      id name  score
##   <dbl> <chr> <dbl>
## 1     2 Irene    76
## 2     4 Chloe    60
```
]
]

---
## Horizontally Merge
如果兩個資料集的索引鍵（key value）名稱不同時，在 `by` argument 中手動指定左右表的索引鍵，寫法為 `by = c("左表索引鍵" = "右表索引鍵")`。

``` r
people <- tibble(
  ID = c(1, 2, 3, 4, 6),
  name = c("Tim", "Irene", "Cindy", "Chloe", "Sandy"))
school <- tribble(
  ~id, ~school,
  3, "NTU",
  4, "NTNU",
  5, "NCCU",
  7, "NTU")

people %>% inner_join(school, by = c("ID" = "id"))
```

```
## # A tibble: 2 × 3
##      ID name  school
##   <dbl> <chr> <chr> 
## 1     3 Cindy NTU   
## 2     4 Chloe NTNU
```

---
## Horizontally Merge
有時候單一索引鍵無法唯一辨別一筆資料。例如 `score` 記錄學生**每一年**的學期成績，`grant` 記錄每個學生**每一年**領到的獎學金金額——同一個 id 在兩張表中都出現多次，此時需同時使用 id, year 作為索引鍵才能辨別唯一資料。

``` r
score <- tribble(
  ~id, ~year, ~name,   ~score,
  1,   2024,  "Tim",   88,
  1,   2025,  "Tim",   91,
  2,   2024,  "Irene", 76,
  2,   2025,  "Irene", 82,
  3,   2024,  "Cindy", 95
)
grant <- tribble(
  ~id, ~year, ~scholarship,
  1,   2024,  8000,
  1,   2025,  12000,
  2,   2024,  5000,
  3,   2025,  10000
)
```

---
## Horizontally Merge
.panelset[
.panel[.panel-name[only key: id]

``` r
score %>% left_join(grant, by = "id")
```

```
## Warning in left_join(., grant, by = "id"): Detected an unexpected many-to-many relationship between `x` and `y`.
## ℹ Row 1 of `x` matches multiple rows in `y`.
## ℹ Row 1 of `y` matches multiple rows in `x`.
## ℹ If a many-to-many relationship is expected, set `relationship =
##   "many-to-many"` to silence this warning.
```

```
## # A tibble: 7 × 6
##      id year.x name  score year.y scholarship
##   <dbl>  <dbl> <chr> <dbl>  <dbl>       <dbl>
## 1     1   2024 Tim      88   2024        8000
## 2     1   2024 Tim      88   2025       12000
## 3     1   2025 Tim      91   2024        8000
## 4     1   2025 Tim      91   2025       12000
## 5     2   2024 Irene    76   2024        5000
## 6     2   2025 Irene    82   2024        5000
## 7     3   2024 Cindy    95   2025       10000
```
]
.panel[.panel-name[both of keys]
將 id, year 一起放進 `by`，**兩個條件同時符合**才算配對成功，資料維持 5 筆。

``` r
score %>% left_join(grant, by = c("id", "year"))
```

```
## # A tibble: 5 × 5
##      id  year name  score scholarship
##   <dbl> <dbl> <chr> <dbl>       <dbl>
## 1     1  2024 Tim      88        8000
## 2     1  2025 Tim      91       12000
## 3     2  2024 Irene    76        5000
## 4     2  2025 Irene    82          NA
## 5     3  2024 Cindy    95          NA
```
沒領獎學金的學生年度（Irene 2025、Cindy 2024）`amount` 為 `NA`。
]

.panel[.panel-name[one difference key name]
一個索引鍵名稱相同、另一個不同，可以**混合寫法**。

``` r
grant2 <- grant %>% rename(yr = year)
score %>% left_join(grant2, by = c("id", "year" = "yr"))
```

---
## Horizontally Merge
前面的範例中 `id` 都是唯一值，所以合併前後筆數不變；但只要有**一邊的索引鍵重複，合併後的資料筆數就會增加**。

``` r
people <- tibble(
  id    = c(1, 2, 3, 4, 6),
  name  = c("Tim", "Irene", "Cindy", "Chloe", "Sandy"),
  score = c(88, 76, 95, 60, 82)
)
club <- tribble(    # 一個人可以參加多個社團
  ~id, ~club,
  1,   "Basketball",
  1,   "Photography",
  2,   "Guitar",
  3,   "Basketball",
  3,   "Debate",
  3,   "Volleyball"
)
```

---
## Horizontally Merge

`people` 原本 5 筆（一人一筆），合併後變成 8 筆：Cindy 參加 3 個社團，會複製成 3 列。

``` r
people %>% left_join(club, by = "id")
```

```
## # A tibble: 8 × 4
##      id name  score club       
##   <dbl> <chr> <dbl> <chr>      
## 1     1 Tim      88 Basketball 
## 2     1 Tim      88 Photography
## 3     2 Irene    76 Guitar     
## 4     3 Cindy    95 Basketball 
## 5     3 Cindy    95 Debate     
## 6     3 Cindy    95 Volleyball 
## 7     4 Chloe    60 <NA>       
## 8     6 Sandy    82 <NA>
```

---
## Practice：Vertically Merge
甲班與乙班的期中考成績分別存在兩張表，乙班因為有補考，所以多了一個 `makeup` 欄位，而且學號欄位的名稱也不一樣。

``` r
class_a <- tribble(
  ~id, ~name,   ~midterm,
  1,   "Tim",   88,
  2,   "Irene", 76,
  3,   "Cindy", 95
)
class_b <- tribble(
  ~stu_id, ~name,   ~midterm, ~makeup,
  4,       "Chloe", 55,       70,
  5,       "Sandy", 82,       NA,
  6,       "Leo",   48,       65
)
```

---
## Practice：Vertically Merge
.panelset[
.panel[.panel-name[question]
#### 1. 直接執行 `rbind(class_a, class_b)` 會發生什麼事？為什麼？

#### 2. 把兩個班級合併成一張完整的成績表 `grade`。

#### 3. 合併後 `makeup` 欄位有幾個 `NA`？分別是誰？
]

.panel[.panel-name[answer1]

``` r
rbind(class_a, class_b)
```

```
## Error in rbind(deparse.level, ...): numbers of columns of arguments do not match
```
`rbind()` 要求欄位數量完全相同。改用 `bind_rows()` 雖然不會報錯，但名稱不一致的欄位會被當成兩個不同變數：

``` r
class_a %>% bind_rows(class_b)
```

```
## # A tibble: 6 × 5
##      id name  midterm stu_id makeup
##   <dbl> <chr>   <dbl>  <dbl>  <dbl>
## 1     1 Tim        88     NA     NA
## 2     2 Irene      76     NA     NA
## 3     3 Cindy      95     NA     NA
## 4    NA Chloe      55      4     70
## 5    NA Sandy      82      5     NA
## 6    NA Leo        48      6     65
```
]

.panel[.panel-name[answer2]

``` r
grade <- class_a %>% bind_rows(class_b %>% rename(id = stu_id))
grade
```

```
## # A tibble: 6 × 4
##      id name  midterm makeup
##   <dbl> <chr>   <dbl>  <dbl>
## 1     1 Tim        88     NA
## 2     2 Irene      76     NA
## 3     3 Cindy      95     NA
## 4     4 Chloe      55     70
## 5     5 Sandy      82     NA
## 6     6 Leo        48     65
```
** 先 `rename()` 讓索引欄位名稱一致，再 `bind_rows()`。**
]

.panel[.panel-name[answer3]

``` r
sum(is.na(grade$makeup))
```

```
## [1] 4
```

``` r
grade %>% filter(is.na(makeup))
```

```
## # A tibble: 4 × 4
##      id name  midterm makeup
##   <dbl> <chr>   <dbl>  <dbl>
## 1     1 Tim        88     NA
## 2     2 Irene      76     NA
## 3     3 Cindy      95     NA
## 4     5 Sandy      82     NA
```

]
]

---
## Practice：Horizontally Merge
`students` 是系上學生名單，`enroll` 是選課紀錄。

``` r
students <- tibble(
  id   = c(1, 2, 3, 4, 5),
  name = c("Tim", "Irene", "Cindy", "Chloe", "Sandy"),
  dept = c("AGEC", "ECON", "AGEC", "STAT", "AGEC")
)
enroll <- tribble(
  ~id, ~course,        ~credit,
  1,   "Econometrics", 3,
  1,   "Statistics",   3,
  2,   "Econometrics", 3,
  3,   "Calculus",     4,
  6,   "Statistics",   3
)
```

---
## Practice：Horizontally Merge 
.panelset[
.panel[.panel-name[question]
#### 1. 保留所有學生資料，將 `enroll` 合併進來存成 `df`。

#### 2. 哪些學生沒有修任何一門課？

#### 3. 列出有修課的學生名單，只保留學生基本資料（不要課程欄位）。

#### 4. 每位學生總共修了幾學分？（沒修課的學生也要出現在結果中）
]

.panel[.panel-name[answer1]

``` r
df <- students %>% left_join(enroll, by = "id"); df
```

```
## # A tibble: 6 × 5
##      id name  dept  course       credit
##   <dbl> <chr> <chr> <chr>         <dbl>
## 1     1 Tim   AGEC  Econometrics      3
## 2     1 Tim   AGEC  Statistics        3
## 3     2 Irene ECON  Econometrics      3
## 4     3 Cindy AGEC  Calculus          4
## 5     4 Chloe STAT  <NA>             NA
## 6     5 Sandy AGEC  <NA>             NA
```
#### 從 5 筆變成 6 筆：Tim 修了兩門課被複製成 2 列。另外 `enroll` 中 id = 6 不在名單裡，`left_join` 直接忽略。
]

.panel[.panel-name[answer2]

``` r
students %>% anti_join(enroll, by = "id")
```

```
## # A tibble: 2 × 3
##      id name  dept 
##   <dbl> <chr> <chr>
## 1     4 Chloe STAT 
## 2     5 Sandy AGEC
```
]

.panel[.panel-name[answer3]

``` r
students %>% semi_join(enroll, by = "id")
```

```
## # A tibble: 3 × 3
##      id name  dept 
##   <dbl> <chr> <chr>
## 1     1 Tim   AGEC 
## 2     2 Irene ECON 
## 3     3 Cindy AGEC
```
** 用 `semi_join()` 而不是 `inner_join()`：Tim 修兩門課也只會出現一次。**
]

.panel[.panel-name[answer4]

``` r
df %>% group_by(id, name) %>%
  summarise(total_credit = sum(credit, na.rm = TRUE), .groups = "drop")
```

```
## # A tibble: 5 × 3
##      id name  total_credit
##   <dbl> <chr>        <dbl>
## 1     1 Tim              6
## 2     2 Irene            3
## 3     3 Cindy            4
## 4     4 Chloe            0
## 5     5 Sandy            0
```
** 要從 `df`（left_join 的結果）算起，沒修課的 Chloe、Sandy 才會留下來變成 0。**
]
]

---
class: center, middle, inverse
### 謝謝
#### 下周見