第13章のRコード

第13章 実証分析の手順

サンプルデータ

Males1980.csv:1980年米国における男性労働者に関するデータ.

Males.csv:1980-1987年米国における男性労働者に関するデータ.

パッケージの読み込み

library(dplyr)
library(ggplot2)
library(patchwork)
library(broom)
library(fixest)
library(modelsummary)

13.1 データを集める

13.1.2 Rの組み込みデータ

data()

help("datasets")

plot(women)

lm(height ~ weight,
   data = women)

13.1.6 データをながめる

malesdata <- readr::read_csv("Males1980.csv")

malesdata
# A tibble: 545 × 3
   school exper  wage
    <dbl> <dbl> <dbl>
 1     14     1 1.20 
 2     13     4 1.68 
 3     12     4 1.52 
 4     12     2 1.89 
 5     12     5 1.95 
 6     10     2 0.259
 7     13     2 1.74 
 8     12     2 1.21 
 9     11     2 0.502
10     10     2 1.01 
# ℹ 535 more rows
malesdata %>% 
  summary()
     school          exper             wage       
 Min.   : 3.00   Min.   : 0.000   Min.   :-1.114  
 1st Qu.:11.00   1st Qu.: 2.000   1st Qu.: 1.165  
 Median :12.00   Median : 3.000   Median : 1.448  
 Mean   :11.77   Mean   : 3.015   Mean   : 1.393  
 3rd Qu.:12.00   3rd Qu.: 4.000   3rd Qu.: 1.742  
 Max.   :16.00   Max.   :11.000   Max.   : 2.822  
View(malesdata)
g1 <- malesdata %>% 
  ggplot(aes(x = wage)) +
  geom_histogram(bins = 15)

g2 <- malesdata %>% 
  ggplot(aes(x = school)) +
  geom_histogram(bins = 15)

g3 <- malesdata %>% 
  ggplot(aes(x = exper)) +
  geom_histogram(bins = 15)

g1 + g2 + g3 + plot_layout(ncol = 1)

13.1.7 データの相関

cor(malesdata)
           school       exper       wage
school  1.0000000 -0.57975653 0.18114299
exper  -0.5797565  1.00000000 0.07689126
wage    0.1811430  0.07689126 1.00000000

13.2 推定結果を検証する

13.2.1 統計的有意性を確認する

malesdata %>% 
  lm(wage ~ school + exper,
     data = .) %>% 
  tidy()
# A tibble: 3 × 5
  term        estimate std.error statistic  p.value
  <chr>          <dbl>     <dbl>     <dbl>    <dbl>
1 (Intercept)  -0.161     0.224     -0.718 4.73e- 1
2 school        0.108     0.0161     6.73  4.22e-11
3 exper         0.0923    0.0170     5.43  8.66e- 8
malesdata %>% 
  lm(wage ~ school + exper,
     data = .) %>% 
  glance() %>% 
  select(r.squared)
# A tibble: 1 × 1
  r.squared
      <dbl>
1    0.0827

13.2.2 モデルを変えてみる

malesdata <- readr::read_csv("Males.csv")

malesdata <- malesdata %>% 
  mutate(black      = 1 * (ethn == "black"),
         industry   = as.factor(industry),
         industry   = relevel(industry, ref = "Manufacturing"),
         occupation = as.factor(occupation),
         occupation = relevel(occupation, ref = "Clerical_and_kindred"))

malesdata %>% 
  filter(year == 1980) %>% 
  lm(wage ~ school + exper + black + health + married +
       union + industry + occupation,
     data = .) %>% 
  tidy() %>% 
  filter(p.value < 0.01)
# A tibble: 8 × 5
  term                                     estimate std.error statistic  p.value
  <chr>                                       <dbl>     <dbl>     <dbl>    <dbl>
1 school                                     0.0847    0.0166      5.11  4.43e-7
2 exper                                      0.0625    0.0170      3.67  2.69e-4
3 marriedyes                                 0.158     0.0596      2.66  8.08e-3
4 unionyes                                   0.210     0.0543      3.87  1.24e-4
5 industryEntertainment                     -0.698     0.204      -3.42  6.74e-4
6 industryMining                            -0.577     0.216      -2.68  7.66e-3
7 industryProfessional_and_Related Service  -0.356     0.101      -3.53  4.44e-4
8 industryTrade                             -0.217     0.0708     -3.07  2.29e-3
malesdata %>% 
  filter(year == 1980) %>% 
  lm(wage ~ school + exper + black + health + married +
       union + industry + occupation,
     data = .) %>% 
  glance() %>% 
  select(r.squared)
# A tibble: 1 × 1
  r.squared
      <dbl>
1     0.210

13.2.3 データを変えてみる

malesdata %>% 
  feols(wage ~ school + exper + black + health + married +
       union + industry + occupation,
     data = .) %>% 
  tidy() %>% 
  filter(p.value < 0.01)
# A tibble: 15 × 5
   term                                    estimate std.error statistic  p.value
   <chr>                                      <dbl>     <dbl>     <dbl>    <dbl>
 1 (Intercept)                               0.418    0.0695       6.02 1.93e- 9
 2 school                                    0.0845   0.00471     17.9  2.31e-69
 3 exper                                     0.0434   0.00281     15.5  1.37e-52
 4 black                                    -0.127    0.0224      -5.66 1.65e- 8
 5 marriedyes                                0.0838   0.0152       5.52 3.55e- 8
 6 unionyes                                  0.182    0.0172      10.6  6.40e-26
 7 industryAgricultural                     -0.378    0.0524      -7.22 6.01e-13
 8 industryBusiness_and_Repair_Service      -0.143    0.0294      -4.87 1.13e- 6
 9 industryConstruction                     -0.0898   0.0297      -3.02 2.51e- 3
10 industryEntertainment                    -0.522    0.0597      -8.74 3.23e-18
11 industryPersonal_Service                 -0.150    0.0571      -2.63 8.57e- 3
12 industryProfessional_and_Related Servi…  -0.319    0.0312     -10.2  3.50e-24
13 industryTrade                            -0.225    0.0210     -10.8  1.16e-26
14 occupationManagers, Officials_and_Prop…   0.136    0.0318       4.28 1.94e- 5
15 occupationProfessional, Technical_and_…   0.158    0.0317       4.97 6.95e- 7

13.2.4 推定手法を変えてみる

malesdata %>% 
  feols(wage ~ school + exper + black + health + married +
          union + industry + occupation | nr + year) %>% 
  tidy() %>% 
  filter(p.value < 0.01)
The variables 'school', 'exper' and 'black' have been removed because of collinearity (see $collin.var).
# A tibble: 4 × 5
  term                                      estimate std.error statistic p.value
  <chr>                                        <dbl>     <dbl>     <dbl>   <dbl>
1 unionyes                                    0.0828    0.0221      3.75 1.95e-4
2 industryEntertainment                      -0.223     0.0685     -3.25 1.23e-3
3 industryTrade                              -0.106     0.0274     -3.87 1.23e-4
4 occupationManagers, Officials_and_Propri…   0.0896    0.0320      2.79 5.38e-3

補足:テーブルの作成方法

記述統計表

malesdata <- readr::read_csv("Males.csv")

malesdata %>% 
  datasummary_skim(histogram = FALSE)
Unique (#) Missing (%) Mean SD Min Median Max
...1 4360 0 2180.5 1258.8 1.0 2180.5 4360.0
nr 545 0 5262.1 3496.1 13.0 4569.0 12548.0
year 8 0 1983.5 2.3 1980.0 1983.5 1987.0
school 13 0 11.8 1.7 3.0 12.0 16.0
exper 19 0 6.5 2.8 0.0 6.0 18.0
wage 3631 0 1.6 0.5 −3.6 1.7 4.1
malesdata <- malesdata %>% 
  mutate(union   = 1 * (union   == "yes"),
         black   = 1 * (ethn    == "black"),
         married = 1 * (married == "yes"),
         health  = 1 * (health  == "yes"))

malesdata %>% 
  select(wage, married, union, health,
         school, exper, black) %>% 
  datasummary_skim(histogram = FALSE,
                   fmt = "%.2f")
Unique (#) Missing (%) Mean SD Min Median Max
wage 3631 0 1.65 0.53 −3.58 1.67 4.05
married 2 0 0.44 0.50 0.00 0.00 1.00
union 2 0 0.24 0.43 0.00 0.00 1.00
health 2 0 0.02 0.13 0.00 0.00 1.00
school 13 0 11.77 1.75 3.00 12.00 16.00
exper 19 0 6.51 2.83 0.00 6.00 18.00
black 2 0 0.12 0.32 0.00 0.00 1.00

推定結果表

m1 <- malesdata %>% 
  feols(wage ~ married + union + health +
          school + exper + black)

m2 <- malesdata %>% 
  feols(wage ~ married + union + health +
          school + exper + black + industry + occupation)

m3 <- malesdata %>% 
  feols(wage ~ married + union + health +
          school + exper + black + industry + occupation | nr + year)
The variables 'school', 'exper' and 'black' have been removed because of collinearity (see $collin.var).
list(m1, m2, m3) %>% 
  modelsummary()
Model 1 Model 2 Model 3
(Intercept) 0.034 0.040
(0.062) (0.084)
married 0.112 0.084 0.051
(0.016) (0.015) (0.021)
union 0.184 0.182 0.083
(0.017) (0.017) (0.022)
health −0.052 −0.033 −0.009
(0.057) (0.054) (0.048)
school 0.103 0.084
(0.004) (0.005)
exper 0.050 0.043
(0.003) (0.003)
black −0.145 −0.127
(0.023) (0.022)
industryBusiness_and_Repair_Service 0.235 0.016
(0.057) (0.072)
industryConstruction 0.288 −0.005
(0.057) (0.079)
industryEntertainment −0.143 −0.148
(0.077) (0.088)
industryFinance 0.410 0.204
(0.064) (0.087)
industryManufacturing 0.378 0.074
(0.052) (0.065)
industryMining 0.479 0.043
(0.075) (0.113)
industryPersonal_Service 0.228 0.064
(0.075) (0.082)
industryProfessional_and_Related Service 0.060 −0.015
(0.058) (0.076)
industryPublic_Administration 0.292 0.056
(0.062) (0.076)
industryTrade 0.153 −0.032
(0.053) (0.065)
industryTransportation 0.411 0.045
(0.058) (0.070)
occupationCraftsmen, Foremen_and_kindred 0.031 0.046
(0.028) (0.031)
occupationFarm_Laborers_and_Foreman −0.059 0.055
(0.078) (0.074)
occupationLaborers_and_farmers −0.067 0.040
(0.032) (0.032)
occupationManagers, Officials_and_Proprietors 0.136 0.090
(0.032) (0.032)
occupationOperatives_and_kindred −0.059 0.039
(0.028) (0.027)
occupationProfessional, Technical_and_kindred 0.158 0.092
(0.032) (0.036)
occupationSales_Workers 0.079 0.027
(0.037) (0.053)
occupationService_Workers −0.077 0.043
(0.030) (0.039)
Num.Obs. 4360 4360 4360
R2 0.184 0.261 0.622
R2 Adj. 0.183 0.257 0.565
R2 Within 0.024
R2 Within Adj. 0.018
AIC 6010.2 5615.2 2684.9
BIC 6061.2 5787.5 2831.6
RMSE 0.48 0.46 0.33
Std.Errors IID IID by: nr
FE: nr X
FE: year X
rows <- tribble(~term, ~`Model 1`, ~`Model 2`, ~`Model 3`,
                "Industry",      "No", "Yes", "Yes",
                "Occupation",    "No", "Yes", "Yes",
                "Individual FE", "No", "No",  "Yes",
                "Year FE",       "No", "No",  "Yes")

attr(rows, "position") <- 15:18

msummary(list(m1, m2, m3),
         coef_omit   = "industry|occupation", 
         coef_rename = c("(Intercept)" = "切片"),
         gof_map     = "nobs",
         add_rows    = rows,
         stars       = TRUE,
         notes       = "備考:対数賃金を被説明変数として線形回帰モデルを最小二乗法により推定して得られた推定値。丸括弧内は標準誤差。")
Model 1 Model 2 Model 3
切片 0.034 0.040
(0.062) (0.084)
married 0.112*** 0.084*** 0.051*
(0.016) (0.015) (0.021)
union 0.184*** 0.182*** 0.083***
(0.017) (0.017) (0.022)
health −0.052 −0.033 −0.009
(0.057) (0.054) (0.048)
school 0.103*** 0.084***
(0.004) (0.005)
exper 0.050*** 0.043***
(0.003) (0.003)
black −0.145*** −0.127***
(0.023) (0.022)
Industry No Yes Yes
Occupation No Yes Yes
Individual FE No No Yes
Year FE No No Yes
Num.Obs. 4360 4360 4360
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001
備考:対数賃金を被説明変数として線形回帰
モデルを最小二乗法により推定して得られた
推定値。丸括弧内は標準誤差。