第12回 交互作用
関西大学総合情報学部
2023-12-21
すぐに実習できるように準備しておきましょう。
Data
フォルダーを作成し、そこにアップロードしましょう。主な説明変数(\(X\))と応答変数(\(Y\))の関係において、\(X\)が\(Y\)に与える影響がその他の変数(\(Z\))の影響を受ける場合
説明変数、調整変数、交差項を投入した回帰モデル
\[ \hat{Y} = \alpha + \beta_1 X + \beta_2 Z + \beta_3 X \cdot Z \]
\[ \hat{Y} = \alpha + (\beta_1 + \beta_3 Z) X + \beta_2 Z \]
調整変数\(Z\)が0、または1の値のみをとるダミー変数の場合(\(Z \in \{0, 1\}\))
\[ \hat{Y} = \alpha + \beta_1 X + \beta_2 Z + \beta_3 X \cdot Z \]
以下のモデルの場合…
\[ \begin{align} \hat{Y} & = 3 + 2 X + 1 Z + 3 X \cdot Z \\ & = 3 + (2 + 3Z) X + 1 Z \end{align} \]
調整変数\(Z\)が無数の値をとる連続変数の場合
\[ \hat{Y} = \alpha + \beta_1 X + \beta_2 Z + \beta_3 X \cdot Z \]
以下のモデルの場合…
\[ \begin{align} \hat{Y} & = 2 + 3 X + 2 Z - 1 X \cdot Z \\ & = 2 + (3 - 1Z) X + 2 Z \end{align} \]
# A tibble: 3,000 × 6
TempKyosan Female Age Satisfaction Interest Ideology
<dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
1 20 1 69 4 4 9
2 20 1 47 1 1 7
3 0 1 37 3 3 11
4 0 0 51 4 3 11
5 20 0 38 2 3 7
6 0 0 71 5 4 11
7 10 0 47 3 3 9
8 0 0 71 4 4 11
9 25 0 75 3 4 9
10 40 1 66 2 3 6
# ℹ 2,990 more rows
変数 | 説明 | 備考 |
---|---|---|
TempKyosan |
日本共産党に対する感情温度 | 高いほど好感 |
Female |
女性ダミー | 0: 男性 / 1: 女性 |
Age |
回答者の年齢 | |
Satisfaction |
政治満足度 | 高いほど満足 |
Interest |
回答者の政治関心 | 高いほど関心あり |
Ideology |
回答者のイデオロギー | 高いほど保守的 |
今回のデータはすべて連続変数扱いとなるため、前処理は不要
Descriptive Statistics
jes_df
N: 3000
Mean Std.Dev Min Max N.Valid
------------------ ------- --------- ------- -------- ---------
TempKyosan 26.88 24.95 0.00 100.00 3000.00
Female 0.50 0.50 0.00 1.00 3000.00
Age 47.34 15.63 18.00 75.00 3000.00
Satisfaction 2.45 1.08 1.00 5.00 3000.00
Interest 2.74 0.83 1.00 4.00 3000.00
Ideology 6.34 2.10 1.00 11.00 3000.00
政治満足度が共産党に対する感情温度に与える影響を調べたい。ただし、この影響は一定ではなく、性別や年齢によって異なるかも知れない。政治満足度が共産党に対する感情温度に与える影響の不均一性を調べるためにはどうすれば良いだろうか。仮説検定に使用する有意水準は5%とする(\(\alpha = 0.05\))。
モデル1
TempKoysan
)Satisfaction
)Female
)\(\leftarrow\) ダミー変数(二値変数)Interest
)、イデオロギー(Ideology
)、年齢(Age
)モデル2
TempKoysan
)Satisfaction
)Age
)\(\leftarrow\) 連続変数Interest
)、イデオロギー(Ideology
)、女性ダミー(Female
)政治満足度(= 説明変数)が共産党に対する感情温度(= 応答変数)に与える影響は性別(= 調整変数)によって変わる。
\[ \widehat{\mbox{TempKyosan}} = \alpha + \beta_1 \mbox{Satisfaction} + \beta_2 \mbox{Female} + \beta_3 \mbox{Interest} + \beta_4 \mbox{Ideology} + \beta_5 \mbox{Age} + \beta_6 (\mbox{Satisfaction} \cdot \mbox{Female}) \]
政治満足度(= 説明変数)が共産党に対する感情温度(= 応答変数)に与える影響は年齢(= 調整変数)によって変わる。
\[ \widehat{\mbox{TempKyosan}} = \alpha + \beta_1 \mbox{Satisfaction} + \beta_2 \mbox{Age} + \beta_3 \mbox{Interest} + \beta_4 \mbox{Ideology} + \beta_5 \mbox{Female} + \beta_6 (\mbox{Satisfaction} \cdot \mbox{Age}) \]
lm()
内の回帰式(formula)に2つの変数を*
でつなぐだけ
A * B
は説明変数としてA
、B
、A
\(\times\)B
が同時に投入することを意味する。係数 | 標準誤差 | t値 | p値 | |
---|---|---|---|---|
切片 | 49.275 | 2.587 | 19.047 | < 0.001 |
政治満足度 | −4.732 | 0.561 | −8.429 | < 0.001 |
女性 | 2.752 | 2.186 | 1.259 | 0.208 |
政治関心 | 0.217 | 0.571 | 0.381 | 0.703 |
イデオロギー | −1.887 | 0.214 | −8.815 | < 0.001 |
年齢 | −0.040 | 0.030 | −1.347 | 0.178 |
政治満足度 * 女性 | 0.902 | 0.817 | 1.104 | 0.270 |
fit1
)fit1
)fit1_pred <- predictions(fit1, newdata = datagrid(Satisfaction = 1:5,
Female = 0:1))
fit1_pred |>
mutate(Female = if_else(Female == 1, "女性", "男性")) |>
ggplot() +
geom_pointrange(aes(x = Satisfaction, y = estimate,
ymin = conf.low, ymax = conf.high, color = Female),
position = position_dodge2(0.5)) +
labs(x = "政治満足度", y = "共産感情温度の予測値と95%信頼区間 (度)",
fill = "", color = "") +
theme_bw() +
theme(legend.position = "bottom")
fit2
)Age
の値 = 20):3.31 - 0.16 \(\times\) 20 = 0.11
Age
の値 = 40):3.31 - 0.16 \(\times\) 40 = -3.09
Age
の値 = 60):3.31 - 0.16 \(\times\) 60 = -6.29
fit2
)fit2_pred <- predictions(fit2, newdata = datagrid(Satisfaction = 1:5,
Age = c(20, 40, 60)))
fit2_pred |>
mutate(Age = factor(Age, levels = c(20, 40, 60),
labels = c("20歳", "40歳", "60歳"))) |>
ggplot(aes(x = Satisfaction)) +
geom_ribbon(aes(y = estimate, ymin = conf.low, ymax = conf.high,
fill = Age), alpha = 0.3) +
geom_line(aes(y = estimate, color = Age), linewidth = 1) +
labs(x = "政治満足度", y = "共産感情温度の予測値と95%信頼区間 (度)",
fill = "", color = "") +
theme_bw() +
theme(legend.position = "bottom")
交互作用: 説明変数が応答変数に与える影響は調整変数の値によって変わる
{marginaleffect}パッケージのslopes()
関数
variables
には説明変数名を指定する(変数名は"
で囲むこと)fit1
の場合:調整変数(Female
)の値が0と1の場合の限界効果
datagrid(Female = c(0, 1))
、またはdatagrid(Female = 0:1)
fit2
の場合:調整変数(Age
)の値が18、19、20、…、75の場合の限界効果
datagrid(Age = 18:75)
estimate
列、\(p\)値はp.value
列、95%信頼区間はconf.low
(下限)とconf.high
(上限)conf_level = 0.9
、0.01ならconf_level = 0.99
を追加fit1
)性別と関係なく、政治満足度は共産感情温度に負の影響を与える。
Term Female Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Satisfaction 0 -4.73 0.561 -8.43 <0.001 54.7 -5.83 -3.63
Satisfaction 1 -3.83 0.611 -6.27 <0.001 31.3 -5.03 -2.63
Columns: rowid, term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, Female, predicted_lo, predicted_hi, predicted, TempKyosan, Satisfaction, Interest, Ideology, Age
Type: response
Female
の値が0の場合
Satisfaction
が1単位上がるとTempKyosan
は約-4.73度下がり、\(\alpha = 0.05\)の水準で統計的に有意である。Female
の値が1の場合
Satisfaction
が1単位上がるとTempKyosan
は約-3.83度下がり、\(\alpha = 0.05\)の水準で統計的に有意である。fit2
)
Term Age Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Satisfaction 18 0.4860 0.913 0.5323 0.595 0.8 -1.30 2.28
Satisfaction 19 0.3292 0.890 0.3700 0.711 0.5 -1.42 2.07
Satisfaction 20 0.1725 0.866 0.1992 0.842 0.2 -1.53 1.87
Satisfaction 21 0.0157 0.843 0.0186 0.985 0.0 -1.64 1.67
Satisfaction 22 -0.1411 0.820 -0.1720 0.863 0.2 -1.75 1.47
--- 48 rows omitted. See ?avg_slopes and ?print.marginaleffects ---
Satisfaction 71 -7.8227 0.724 -10.8071 <0.001 88.0 -9.24 -6.40
Satisfaction 72 -7.9795 0.746 -10.7029 <0.001 86.4 -9.44 -6.52
Satisfaction 73 -8.1362 0.768 -10.5965 <0.001 84.7 -9.64 -6.63
Satisfaction 74 -8.2930 0.791 -10.4875 <0.001 83.1 -9.84 -6.74
Satisfaction 75 -8.4498 0.813 -10.3960 <0.001 81.7 -10.04 -6.86
Columns: rowid, term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, Age, predicted_lo, predicted_hi, predicted, TempKyosan, Satisfaction, Interest, Ideology, Female
Type: response
Age
が18の場合、Satisfaction
はTempKyosan
に影響を与えているとは言えない(\(p \geq 0.05\)のため)。Age
が19の場合、Satisfaction
はTempKyosan
に影響を与えているとは言えない(\(p \geq 0.05\)のため)。Age
が75の場合、Satisfaction
が1単位上がるとTempKyosan
は約-8.4度下がり、\(\alpha = 0.05\)の水準で統計的に有意である。print()
内にtopn = Inf
を追加する。
Term Age Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Satisfaction 18 0.4860 0.913 0.5323 0.59453 0.8 -1.30 2.2756
Satisfaction 19 0.3292 0.890 0.3700 0.71142 0.5 -1.42 2.0736
Satisfaction 20 0.1725 0.866 0.1992 0.84214 0.2 -1.53 1.8700
Satisfaction 21 0.0157 0.843 0.0186 0.98513 0.0 -1.64 1.6681
Satisfaction 22 -0.1411 0.820 -0.1720 0.86345 0.2 -1.75 1.4664
Satisfaction 23 -0.2978 0.797 -0.3735 0.70878 0.5 -1.86 1.2650
Satisfaction 24 -0.4546 0.775 -0.5866 0.55747 0.8 -1.97 1.0643
Satisfaction 25 -0.6114 0.753 -0.8121 0.41676 1.3 -2.09 0.8642
Satisfaction 26 -0.7681 0.731 -1.0506 0.29346 1.8 -2.20 0.6649
Satisfaction 27 -0.9249 0.709 -1.3038 0.19230 2.4 -2.32 0.4655
Satisfaction 28 -1.0817 0.688 -1.5718 0.11599 3.1 -2.43 0.2671
Satisfaction 29 -1.2384 0.667 -1.8567 0.06336 4.0 -2.55 0.0689
Satisfaction 30 -1.3952 0.646 -2.1589 0.03086 5.0 -2.66 -0.1286
Satisfaction 31 -1.5520 0.627 -2.4754 0.01331 6.2 -2.78 -0.3232
Satisfaction 32 -1.7087 0.608 -2.8102 0.00495 7.7 -2.90 -0.5170
Satisfaction 33 -1.8655 0.588 -3.1710 0.00152 9.4 -3.02 -0.7124
Satisfaction 34 -2.0223 0.570 -3.5507 < 0.001 11.3 -3.14 -0.9060
Satisfaction 35 -2.1790 0.552 -3.9465 < 0.001 13.6 -3.26 -1.0969
Satisfaction 36 -2.3358 0.535 -4.3637 < 0.001 16.3 -3.38 -1.2867
Satisfaction 37 -2.4926 0.519 -4.8017 < 0.001 19.3 -3.51 -1.4752
Satisfaction 38 -2.6493 0.504 -5.2573 < 0.001 22.7 -3.64 -1.6616
Satisfaction 39 -2.8061 0.490 -5.7325 < 0.001 26.6 -3.77 -1.8467
Satisfaction 40 -2.9629 0.476 -6.2219 < 0.001 30.9 -3.90 -2.0295
Satisfaction 41 -3.1196 0.464 -6.7230 < 0.001 35.7 -4.03 -2.2102
Satisfaction 42 -3.2764 0.453 -7.2319 < 0.001 40.9 -4.16 -2.3885
Satisfaction 43 -3.4332 0.443 -7.7476 < 0.001 46.6 -4.30 -2.5647
Satisfaction 44 -3.5900 0.435 -8.2483 < 0.001 52.5 -4.44 -2.7369
Satisfaction 45 -3.7467 0.428 -8.7458 < 0.001 58.6 -4.59 -2.9071
Satisfaction 46 -3.9035 0.424 -9.2165 < 0.001 64.8 -4.73 -3.0734
Satisfaction 47 -4.0603 0.420 -9.6765 < 0.001 71.2 -4.88 -3.2379
Satisfaction 48 -4.2170 0.418 -10.0965 < 0.001 77.2 -5.04 -3.3984
Satisfaction 49 -4.3738 0.417 -10.4787 < 0.001 82.9 -5.19 -3.5557
Satisfaction 50 -4.5306 0.419 -10.8114 < 0.001 88.1 -5.35 -3.7092
Satisfaction 51 -4.6873 0.422 -11.1103 < 0.001 92.9 -5.51 -3.8604
Satisfaction 52 -4.8441 0.427 -11.3534 < 0.001 96.8 -5.68 -4.0078
Satisfaction 53 -5.0009 0.433 -11.5539 < 0.001 100.2 -5.85 -4.1525
Satisfaction 54 -5.1576 0.441 -11.7037 < 0.001 102.7 -6.02 -4.2939
Satisfaction 55 -5.3144 0.450 -11.8161 < 0.001 104.6 -6.20 -4.4329
Satisfaction 56 -5.4712 0.461 -11.8789 < 0.001 105.7 -6.37 -4.5685
Satisfaction 57 -5.6279 0.472 -11.9191 < 0.001 106.4 -6.55 -4.7025
Satisfaction 58 -5.7847 0.485 -11.9240 < 0.001 106.5 -6.74 -4.8339
Satisfaction 59 -5.9415 0.499 -11.9041 < 0.001 106.1 -6.92 -4.9632
Satisfaction 60 -6.0982 0.514 -11.8646 < 0.001 105.4 -7.11 -5.0908
Satisfaction 61 -6.2550 0.530 -11.7955 < 0.001 104.3 -7.29 -5.2157
Satisfaction 62 -6.4118 0.547 -11.7245 < 0.001 103.0 -7.48 -5.3399
Satisfaction 63 -6.5685 0.564 -11.6421 < 0.001 101.6 -7.67 -5.4627
Satisfaction 64 -6.7253 0.582 -11.5471 < 0.001 100.0 -7.87 -5.5838
Satisfaction 65 -6.8821 0.601 -11.4468 < 0.001 98.4 -8.06 -5.7037
Satisfaction 66 -7.0388 0.620 -11.3461 < 0.001 96.7 -8.25 -5.8229
Satisfaction 67 -7.1956 0.640 -11.2416 < 0.001 95.0 -8.45 -5.9411
Satisfaction 68 -7.3524 0.661 -11.1268 < 0.001 93.1 -8.65 -6.0573
Satisfaction 69 -7.5092 0.682 -11.0140 < 0.001 91.3 -8.85 -6.1729
Satisfaction 70 -7.6659 0.702 -10.9129 < 0.001 89.7 -9.04 -6.2891
Satisfaction 71 -7.8227 0.724 -10.8071 < 0.001 88.0 -9.24 -6.4040
Satisfaction 72 -7.9795 0.746 -10.7029 < 0.001 86.4 -9.44 -6.5182
Satisfaction 73 -8.1362 0.768 -10.5965 < 0.001 84.7 -9.64 -6.6313
Satisfaction 74 -8.2930 0.791 -10.4875 < 0.001 83.1 -9.84 -6.7432
Satisfaction 75 -8.4498 0.813 -10.3960 < 0.001 81.7 -10.04 -6.8567
Columns: rowid, term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, Age, predicted_lo, predicted_hi, predicted, TempKyosan, Satisfaction, Interest, Ideology, Female
Type: response
Satisfaction
)は共産感情温度(TempKyosan
)に影響を与えるとは言えない。fit1_ame
)fit2_ame
)geom_hline(yintercept = 0)
レイヤーを追加fit1_ame
やfit2_ame
)と図両方を見る必要がある。
fit1_ame
の例
Term Female Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Satisfaction 0 -4.73 0.561 -8.43 <0.001 54.7 -5.83 -3.63
Satisfaction 1 -3.83 0.611 -6.27 <0.001 31.3 -5.03 -2.63
Columns: rowid, term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, Female, predicted_lo, predicted_hi, predicted, TempKyosan, Satisfaction, Interest, Ideology, Age
Type: response
fit1_ame
の中身と図、両方を見て解釈する。
fit2_ame
の例
Term Age Estimate Std. Error z Pr(>|z|) S 2.5 % 97.5 %
Satisfaction 18 0.4860 0.913 0.5323 0.59453 0.8 -1.30 2.2756
Satisfaction 19 0.3292 0.890 0.3700 0.71142 0.5 -1.42 2.0736
Satisfaction 20 0.1725 0.866 0.1992 0.84214 0.2 -1.53 1.8700
Satisfaction 21 0.0157 0.843 0.0186 0.98513 0.0 -1.64 1.6681
Satisfaction 22 -0.1411 0.820 -0.1720 0.86345 0.2 -1.75 1.4664
Satisfaction 23 -0.2978 0.797 -0.3735 0.70878 0.5 -1.86 1.2650
Satisfaction 24 -0.4546 0.775 -0.5866 0.55747 0.8 -1.97 1.0643
Satisfaction 25 -0.6114 0.753 -0.8121 0.41676 1.3 -2.09 0.8642
Satisfaction 26 -0.7681 0.731 -1.0506 0.29346 1.8 -2.20 0.6649
Satisfaction 27 -0.9249 0.709 -1.3038 0.19230 2.4 -2.32 0.4655
Satisfaction 28 -1.0817 0.688 -1.5718 0.11599 3.1 -2.43 0.2671
Satisfaction 29 -1.2384 0.667 -1.8567 0.06336 4.0 -2.55 0.0689
Satisfaction 30 -1.3952 0.646 -2.1589 0.03086 5.0 -2.66 -0.1286
Satisfaction 31 -1.5520 0.627 -2.4754 0.01331 6.2 -2.78 -0.3232
Satisfaction 32 -1.7087 0.608 -2.8102 0.00495 7.7 -2.90 -0.5170
Satisfaction 33 -1.8655 0.588 -3.1710 0.00152 9.4 -3.02 -0.7124
Satisfaction 34 -2.0223 0.570 -3.5507 < 0.001 11.3 -3.14 -0.9060
Satisfaction 35 -2.1790 0.552 -3.9465 < 0.001 13.6 -3.26 -1.0969
Satisfaction 36 -2.3358 0.535 -4.3637 < 0.001 16.3 -3.38 -1.2867
Satisfaction 37 -2.4926 0.519 -4.8017 < 0.001 19.3 -3.51 -1.4752
Satisfaction 38 -2.6493 0.504 -5.2573 < 0.001 22.7 -3.64 -1.6616
Satisfaction 39 -2.8061 0.490 -5.7325 < 0.001 26.6 -3.77 -1.8467
Satisfaction 40 -2.9629 0.476 -6.2219 < 0.001 30.9 -3.90 -2.0295
Satisfaction 41 -3.1196 0.464 -6.7230 < 0.001 35.7 -4.03 -2.2102
Satisfaction 42 -3.2764 0.453 -7.2319 < 0.001 40.9 -4.16 -2.3885
Satisfaction 43 -3.4332 0.443 -7.7476 < 0.001 46.6 -4.30 -2.5647
Satisfaction 44 -3.5900 0.435 -8.2483 < 0.001 52.5 -4.44 -2.7369
Satisfaction 45 -3.7467 0.428 -8.7458 < 0.001 58.6 -4.59 -2.9071
Satisfaction 46 -3.9035 0.424 -9.2165 < 0.001 64.8 -4.73 -3.0734
Satisfaction 47 -4.0603 0.420 -9.6765 < 0.001 71.2 -4.88 -3.2379
Satisfaction 48 -4.2170 0.418 -10.0965 < 0.001 77.2 -5.04 -3.3984
Satisfaction 49 -4.3738 0.417 -10.4787 < 0.001 82.9 -5.19 -3.5557
Satisfaction 50 -4.5306 0.419 -10.8114 < 0.001 88.1 -5.35 -3.7092
Satisfaction 51 -4.6873 0.422 -11.1103 < 0.001 92.9 -5.51 -3.8604
Satisfaction 52 -4.8441 0.427 -11.3534 < 0.001 96.8 -5.68 -4.0078
Satisfaction 53 -5.0009 0.433 -11.5539 < 0.001 100.2 -5.85 -4.1525
Satisfaction 54 -5.1576 0.441 -11.7037 < 0.001 102.7 -6.02 -4.2939
Satisfaction 55 -5.3144 0.450 -11.8161 < 0.001 104.6 -6.20 -4.4329
Satisfaction 56 -5.4712 0.461 -11.8789 < 0.001 105.7 -6.37 -4.5685
Satisfaction 57 -5.6279 0.472 -11.9191 < 0.001 106.4 -6.55 -4.7025
Satisfaction 58 -5.7847 0.485 -11.9240 < 0.001 106.5 -6.74 -4.8339
Satisfaction 59 -5.9415 0.499 -11.9041 < 0.001 106.1 -6.92 -4.9632
Satisfaction 60 -6.0982 0.514 -11.8646 < 0.001 105.4 -7.11 -5.0908
Satisfaction 61 -6.2550 0.530 -11.7955 < 0.001 104.3 -7.29 -5.2157
Satisfaction 62 -6.4118 0.547 -11.7245 < 0.001 103.0 -7.48 -5.3399
Satisfaction 63 -6.5685 0.564 -11.6421 < 0.001 101.6 -7.67 -5.4627
Satisfaction 64 -6.7253 0.582 -11.5471 < 0.001 100.0 -7.87 -5.5838
Satisfaction 65 -6.8821 0.601 -11.4468 < 0.001 98.4 -8.06 -5.7037
Satisfaction 66 -7.0388 0.620 -11.3461 < 0.001 96.7 -8.25 -5.8229
Satisfaction 67 -7.1956 0.640 -11.2416 < 0.001 95.0 -8.45 -5.9411
Satisfaction 68 -7.3524 0.661 -11.1268 < 0.001 93.1 -8.65 -6.0573
Satisfaction 69 -7.5092 0.682 -11.0140 < 0.001 91.3 -8.85 -6.1729
Satisfaction 70 -7.6659 0.702 -10.9129 < 0.001 89.7 -9.04 -6.2891
Satisfaction 71 -7.8227 0.724 -10.8071 < 0.001 88.0 -9.24 -6.4040
Satisfaction 72 -7.9795 0.746 -10.7029 < 0.001 86.4 -9.44 -6.5182
Satisfaction 73 -8.1362 0.768 -10.5965 < 0.001 84.7 -9.64 -6.6313
Satisfaction 74 -8.2930 0.791 -10.4875 < 0.001 83.1 -9.84 -6.7432
Satisfaction 75 -8.4498 0.813 -10.3960 < 0.001 81.7 -10.04 -6.8567
Columns: rowid, term, estimate, std.error, statistic, p.value, s.value, conf.low, conf.high, Age, predicted_lo, predicted_hi, predicted, TempKyosan, Satisfaction, Interest, Ideology, Female
Type: response
fit2_ame |>
ggplot(aes(x = Age)) +
geom_ribbon(aes(ymin = conf.low, ymax = conf.high), fill = "gray80") +
geom_hline(yintercept = 0) +
geom_line(aes(y = estimate), linewidth = 1) +
labs(x = "年齢", y = "政治満足度が共産感情温度に\n与える影響と95%信頼区間") +
scale_x_continuous(breaks = c(18, 20, 30, 40, 50, 60, 70, 75),
labels = c(18, 20, 30, 40, 50, 60, 70, 75)) +
theme_bw(base_size = 12)
fit2_ame
の中身と図、両方を見て解釈する。