如何在R中仅用逻辑回归截距与系数绘制列线图？

方法一：构造lrm对象使用rms::nomogram() #

rms包的nomogram()确实需要基于lrm对象，但你可以通过拟合一个假模型并替换系数的方式实现，无需真实数据集：

# 加载包
library(rms)

# 定义你的系数
coef_intercept <- -12.364
coef_X <- 0.826
coef_Y <- 0.426
coef_Z <- 5.111

# 创建假数据集（匹配变量类型和范围）
set.seed(123)
fake_data <- data.frame(
  X = runif(100, 0, 20),    # 连续变量X：0-20
  Y = runif(100, 0, 20),    # 连续变量Y：0-20
  Z = factor(sample(c(0,1), 100, replace = TRUE)),  # 二分类变量Z
  y = factor(sample(c(0,1), 100, replace = TRUE))   # 假二分类因变量
)

# 拟合临时lrm模型
fake_model <- lrm(y ~ X + Y + Z, data = fake_data)

# 替换模型系数为真实值
fake_model$coefficients <- c(coef_intercept, coef_X, coef_Y, coef_Z)

# 调整变量范围（确保列线图显示正确区间）
fake_model$Design$limits <- list(X = c(0,20), Y = c(0,20), Z = c(0,1))

# 绘制列线图
nomogram(fake_model,
         fun = plogis,          # 逻辑回归概率转换函数
         funlabel = "Probability",  # 概率轴标签
         lp = FALSE)            # 隐藏线性预测值轴（可选）

注意：临时模型的统计量（如p值、R²）是无效的，但不影响列线图的生成，因为列线图仅依赖系数和变量范围。

方法二：用ggplot2手动绘制列线图 #

如果你不想依赖rms包，可以用ggplot2手动构建列线图，灵活性更高：

library(ggplot2)
library(dplyr)

# 定义系数
coef_intercept <- -12.364
coef_X <- 0.826
coef_Y <- 0.426
coef_Z <- 5.111

# 1. 生成变量序列与得分映射
# 以X的最大得分（0.826*20）为基准，将得分缩放至0-100区间
score_scale <- 100 / (coef_X * 20)

# 连续变量X、Y的得分数据
x_data <- tibble(
  var = "X",
  value = seq(0, 20, length.out = 100),
  score = coef_X * value * score_scale
)

y_data <- tibble(
  var = "Y",
  value = seq(0, 20, length.out = 100),
  score = coef_Y * value * score_scale
)

# 二分类变量Z的得分数据
z_data <- tibble(
  var = "Z",
  value = factor(c(0,1)),
  score = coef_Z * as.numeric(value) * score_scale
)

var_scores <- bind_rows(x_data, y_data, z_data)

# 2. 生成总得分对应的概率数据
total_score_seq <- seq(0, max(var_scores$score)*3, length.out = 100)
prob_data <- tibble(
  var = "Probability",
  score = total_score_seq,
  value = plogis( (total_score_seq / score_scale) + coef_intercept )
)

# 3. 绘制列线图
ggplot() +
  # 绘制连续变量的得分-值连线
  geom_line(data = filter(var_scores, var %in% c("X", "Y")),
            aes(x = score, y = value), color = "black") +
  # 绘制二分类变量的得分点
  geom_point(data = z_data, aes(x = score, y = as.numeric(value)), size = 3, color = "black") +
  # 绘制概率轴连线
  geom_line(data = prob_data, aes(x = score, y = value), color = "black") +
  # 自定义y轴刻度与标签（对应各个变量）
  scale_y_continuous(
    breaks = c(0,5,10,15,20, 0,1, 0,0.2,0.4,0.6,0.8,1),
    labels = c("0","5","10","15","20", "0","1", "0","0.2","0.4","0.6","0.8","1")
  ) +
  # 自定义x轴（得分轴）
  scale_x_continuous(breaks = seq(0, 300, 50)) +
  # 添加变量标签
  annotate("text", x = -20, y = 10, label = "X", size = 5) +
  annotate("text", x = -20, y = 35, label = "Y", size = 5) +
  annotate("text", x = -20, y = 60, label = "Z", size = 5) +
  annotate("text", x = -20, y = 85, label = "Probability", size = 5) +
  # 调整主题与布局
  theme_minimal() +
  theme(
    axis.title = element_blank(),
    panel.grid = element_blank(),
    plot.margin = margin(10,10,10,60)
  )

这个方法需要手动调整刻度和布局，但完全不需要依赖模型对象，适合自定义需求。

内容的提问来源于stack exchange，提问作者ava