如何将stm包searchK生成的合并图拆分为4个带原标签的独立图？

刚好我之前也处理过类似的需求，stm包的searchK()默认生成的2x2合并图没法直接拆分，但我们可以手动提取返回结果里的原始数据，自己绘制每个带对应标签的独立图表，具体步骤如下：

1. 先明确合并图里的四个子图对应的数据 #

searchK()返回的df1对象里，df1$results包含了所有绘图需要的核心数据：

• 语义连贯性（Semantic Coherence）：df1$results$semcoh
• 排他性（Exclusivity）：df1$results$exclus
• 持有可能性（Heldout Likelihood）：df1$results$heldout
• 残差（Residuals）：df1$results$residuals
• 对应的K值：df1$results$K

2. 用基础R绘图生成独立子图 #

先提取所需数据，再逐个绘制每个指标的独立图：

# 从df1中提取绘图数据
k_values <- df1$results$K
sem_coh <- df1$results$semcoh
exclusivity <- df1$results$exclus
heldout_likelihood <- df1$results$heldout
residuals <- df1$results$residuals

# 1. 语义连贯性图（对应原合并图左上）
plot(k_values, sem_coh, type = "b", pch = 16, 
     xlab = "Number of Topics (K)", ylab = "Semantic Coherence",
     main = "Semantic Coherence by Number of Topics")

# 2. 排他性图（对应原合并图右上）
plot(k_values, exclusivity, type = "b", pch = 16, 
     xlab = "Number of Topics (K)", ylab = "Exclusivity",
     main = "Exclusivity by Number of Topics")

# 3. 持有可能性图（对应原合并图左下）
plot(k_values, heldout_likelihood, type = "b", pch = 16, 
     xlab = "Number of Topics (K)", ylab = "Heldout Likelihood",
     main = "Heldout Likelihood by Number of Topics")

# 4. 残差图（对应原合并图右下）
plot(k_values, residuals, type = "b", pch = 16, 
     xlab = "Number of Topics (K)", ylab = "Residuals",
     main = "Residuals by Number of Topics")

3. （可选）用ggplot2绘制更美观的独立子图 #

如果你偏好ggplot2的风格，可以把数据整理成数据框后逐个绘制：

library(ggplot2)

# 整理成适合ggplot2的长格式数据框
plot_data <- data.frame(
  K = rep(k_values, 4),
  Value = c(sem_coh, exclusivity, heldout_likelihood, residuals),
  Metric = rep(c("Semantic Coherence", "Exclusivity", "Heldout Likelihood", "Residuals"), each = length(k_values))
)

# 绘制语义连贯性图
ggplot(plot_data[plot_data$Metric == "Semantic Coherence", ], aes(x = K, y = Value)) +
  geom_line(color = "#2c3e50", linewidth = 1) +
  geom_point(color = "#e74c3c", size = 3) +
  labs(x = "Number of Topics (K)", y = "Semantic Coherence", title = "Semantic Coherence by Number of Topics") +
  theme_minimal()

# 绘制排他性图
ggplot(plot_data[plot_data$Metric == "Exclusivity", ], aes(x = K, y = Value)) +
  geom_line(color = "#2c3e50", linewidth = 1) +
  geom_point(color = "#e74c3c", size = 3) +
  labs(x = "Number of Topics (K)", y = "Exclusivity", title = "Exclusivity by Number of Topics") +
  theme_minimal()

# 绘制持有可能性图
ggplot(plot_data[plot_data$Metric == "Heldout Likelihood", ], aes(x = K, y = Value)) +
  geom_line(color = "#2c3e50", linewidth = 1) +
  geom_point(color = "#e74c3c", size = 3) +
  labs(x = "Number of Topics (K)", y = "Heldout Likelihood", title = "Heldout Likelihood by Number of Topics") +
  theme_minimal()

# 绘制残差图
ggplot(plot_data[plot_data$Metric == "Residuals", ], aes(x = K, y = Value)) +
  geom_line(color = "#2c3e50", linewidth = 1) +
  geom_point(color = "#e74c3c", size = 3) +
  labs(x = "Number of Topics (K)", y = "Residuals", title = "Residuals by Number of Topics") +
  theme_minimal()

这样就能得到四个和原合并图标签一致的独立图表啦，每个图都可以单独保存或调整样式。

内容的提问来源于stack exchange，提问作者Nathalie