对重复处理而不是参数执行 pca

Question

我有一个数据集，其中第 1 列包含治疗名称，其余列包含这些治疗的值，每个治疗有三个重复。为了说明，我使用 iris 数据集创建了模拟数据集，如下所示：

df <- read.table(text = '"Treatment" "Sepal.Length" "Sepal.Width" "Petal.Length" "Petal.Width"
"treatment_a" 5.1 3.5 1.4 0.2
"treatment_a" 4.9 3 1.4 0.2
"treatment_a" 4.7 3.2 1.3 0.2
"treatment_b" 4.6 3.1 1.5 0.2
"treatment_b" 5 3.6 1.4 0.2
"treatment_b" 5.4 3.9 1.7 0.4
"treatment_c" 4.6 3.4 1.4 0.3
"treatment_c" 5 3.4 1.5 0.2
"treatment_c" 4.4 2.9 1.4 0.2
"treatment_d" 4.9 3.1 1.5 0.1
"treatment_d" 5.4 3.7 1.5 0.2
"treatment_d" 4.8 3.4 1.6 0.2
"treatment_e" 4.8 3 1.4 0.1
"treatment_e" 4.3 3 1.1 0.1
"treatment_e" 5.8 4 1.2 0.2
"treatment_f" 5.7 4.4 1.5 0.4
"treatment_f" 5.4 3.9 1.3 0.4
"treatment_f" 5.1 3.5 1.4 0.3
"treatment_g" 5.7 3.8 1.7 0.3
"treatment_g" 5.1 3.8 1.5 0.3
"treatment_g" 5.4 3.4 1.7 0.2
"treatment_h" 5.1 3.7 1.5 0.4
"treatment_h" 4.6 3.6 1 0.2
"treatment_h" 5.1 3.3 1.7 0.5', header = TRUE)

我想使用 R 在此数据集上执行 pca，其方式是将重复处理而不是变量绘制在图上，处理名称也应在图上标记。 我在 stackoverflow 上寻找了类似的问题，但没有找到与我的问题类似的问题。

Answer 1

原始回复

您是否希望分别在 x 轴和 y 轴上绘制第一个和第二个主成分的散点图？然后你想用治疗标记点吗？如果是这样，你可以试一试。我正在使用ggplot2 包。

我还为锅添加了色彩美感。如果您不想要它，请随意删除该部分。

df <- read.table(text = '"Treatment" "Sepal.Length" "Sepal.Width" "Petal.Length" "Petal.Width"
"treatment_a" 5.1 3.5 1.4 0.2
"treatment_a" 4.9 3 1.4 0.2
"treatment_a" 4.7 3.2 1.3 0.2
"treatment_b" 4.6 3.1 1.5 0.2
"treatment_b" 5 3.6 1.4 0.2
"treatment_b" 5.4 3.9 1.7 0.4
"treatment_c" 4.6 3.4 1.4 0.3
"treatment_c" 5 3.4 1.5 0.2
"treatment_c" 4.4 2.9 1.4 0.2
"treatment_d" 4.9 3.1 1.5 0.1
"treatment_d" 5.4 3.7 1.5 0.2
"treatment_d" 4.8 3.4 1.6 0.2
"treatment_e" 4.8 3 1.4 0.1
"treatment_e" 4.3 3 1.1 0.1
"treatment_e" 5.8 4 1.2 0.2
"treatment_f" 5.7 4.4 1.5 0.4
"treatment_f" 5.4 3.9 1.3 0.4
"treatment_f" 5.1 3.5 1.4 0.3
"treatment_g" 5.7 3.8 1.7 0.3
"treatment_g" 5.1 3.8 1.5 0.3
"treatment_g" 5.4 3.4 1.7 0.2
"treatment_h" 5.1 3.7 1.5 0.4
"treatment_h" 4.6 3.6 1 0.2
"treatment_h" 5.1 3.3 1.7 0.5', header = TRUE)

# run principle components, ignore first column
pr <- prcomp(df[, 2:5])

# run predict to get the first and second principle components
pr_pred <- predict(pr)

# put this into a data frame so we can use ggplot
df2 <- data.frame(Treatment = df$Treatment,
                  pr_pred[, 1:2])

library(ggplot2)

ggplot(data = df2, aes(x = PC1, y = PC2, 
                       colour = Treatment, 
                       label = Treatment)) + 
    geom_text()

添加了省略号

要添加这些，我们必须更改类别的数量。我们三个一起去。希望在您的实际数据集中，有足够的数据来绘制您正在寻找的椭圆。

df_mod <- read.table(text = '"Treatment" "Sepal.Length" "Sepal.Width" "Petal.Length" "Petal.Width"
"treatment_a" 5.1 3.5 1.4 0.2
                 "treatment_a" 4.9 3 1.4 0.2
                 "treatment_a" 4.7 3.2 1.3 0.2
                 "treatment_b" 4.6 3.1 1.5 0.2
                 "treatment_b" 5 3.6 1.4 0.2
                 "treatment_b" 5.4 3.9 1.7 0.4
                 "treatment_c" 4.6 3.4 1.4 0.3
                 "treatment_c" 5 3.4 1.5 0.2
                 "treatment_c" 4.4 2.9 1.4 0.2
                 "treatment_a" 4.9 3.1 1.5 0.1
                 "treatment_a" 5.4 3.7 1.5 0.2
                 "treatment_a" 4.8 3.4 1.6 0.2
                 "treatment_b" 4.8 3 1.4 0.1
                 "treatment_b" 4.3 3 1.1 0.1
                 "treatment_b" 5.8 4 1.2 0.2
                 "treatment_c" 5.7 4.4 1.5 0.4
                 "treatment_c" 5.4 3.9 1.3 0.4
                 "treatment_c" 5.1 3.5 1.4 0.3
                 "treatment_a" 5.7 3.8 1.7 0.3
                 "treatment_a" 5.1 3.8 1.5 0.3
                 "treatment_b" 5.4 3.4 1.7 0.2
                 "treatment_b" 5.1 3.7 1.5 0.4
                 "treatment_c" 4.6 3.6 1 0.2
                 "treatment_c" 5.1 3.3 1.7 0.5', header = TRUE)


pr_mod <- prcomp(df_mod[, 2:5])
pr_pred_mod <- predict(pr_mod)

df2_mod <- data.frame(Treatment = df_mod$Treatment,
                  pr_pred_mod[, 1:2])

ggplot(data = df2_mod, aes(x = PC1, y = PC2, 
                       colour = Treatment, 
                       label = Treatment)) + 
    geom_text() + 
    stat_ellipse(show.legend = FALSE)