如何在基础R（非ggplot）中为多图添加指数与二次多项式趋势线

基础R完全支持绘制非线性趋势线，无需依赖仅能生成直线的abline()，核心思路是先拟合对应模型，生成连续时间序列的预测值，再用lines()绘制平滑趋势线。以下是针对你的数据的具体实现代码和说明：

核心步骤 #

1. 拟合对应模型：
   • 指标S：用nls()拟合指数衰减模型（或对数转换后用lm()，稳定性更高）
   • 指标k、Bg、Br：用lm()拟合二次多项式模型
2. 生成连续时间序列的预测值，确保趋势线平滑
3. 在散点图上用lines()叠加趋势线

完整代码示例 #

1. 模拟示例数据（替换为你的真实数据） #

set.seed(123)
time <- seq(0, 12, by=2)
treatment <- rep(c("Control", "Grazing"), each=length(time))
S <- c(100, 80, 65, 52, 42, 35, 30, 22, 18, 14, 11, 9) + rnorm(12, 0, 3)
k <- c(0.1, 0.15, 0.22, 0.28, 0.3, 0.27, 0.12, 0.18, 0.25, 0.32, 0.35, 0.33) + rnorm(12, 0, 0.02)
Bg <- k * 1.2 + rnorm(12, 0, 0.03)
Br <- k * 0.8 + rnorm(12, 0, 0.02)
decomp_data <- data.frame(time, treatment, S, k, Bg, Br)

2. 绘制多图并添加趋势线 #

# 设置2行2列的多图布局，调整边距避免标题/坐标轴重叠
par(mfrow = c(2, 2), mar = c(4, 4, 2, 1))

# --- 指标S：指数衰减趋势 ---
plot(S ~ time, data = decomp_data, 
     col = ifelse(treatment == "Control", "blue", "red"),
     pch = 16, main = "S: 指数衰减趋势", 
     xlab = "时间", ylab = "S值")
# 拟合对照样地模型（两种方法选其一）
# 方法1：nls直接拟合指数模型
ctrl_S <- nls(S ~ a*exp(-b*time), 
              data = decomp_data[decomp_data$treatment == "Control",],
              start = list(a = 100, b = 0.1))
# 方法2：对数转换后用线性模型（拟合更稳定）
# ctrl_S_lm <- lm(log(S) ~ time, data = decomp_data[decomp_data$treatment == "Control",])

# 生成连续时间预测序列（保证趋势线平滑）
pred_time <- seq(min(time), max(time), length.out = 100)
# 对应方法1的预测
pred_ctrl_S <- predict(ctrl_S, newdata = data.frame(time = pred_time))
# 对应方法2的预测
# pred_ctrl_S <- exp(predict(ctrl_S_lm, newdata = data.frame(time = pred_time)))
lines(pred_time, pred_ctrl_S, col = "blue", lwd = 2)

# 拟合并绘制放牧样地趋势线
graz_S <- nls(S ~ a*exp(-b*time), 
              data = decomp_data[decomp_data$treatment == "Grazing",],
              start = list(a = 100, b = 0.1))
pred_graz_S <- predict(graz_S, newdata = data.frame(time = pred_time))
lines(pred_time, pred_graz_S, col = "red", lwd = 2)
legend("topright", legend = c("对照", "放牧"), 
       col = c("blue", "red"), pch = 16, lwd = 2)

# --- 指标k：二次多项式趋势 ---
plot(k ~ time, data = decomp_data, 
     col = ifelse(treatment == "Control", "blue", "red"),
     pch = 16, main = "k: 二次多项式趋势", 
     xlab = "时间", ylab = "k值")
# 拟合对照样地二次模型
ctrl_k <- lm(k ~ poly(time, 2, raw = TRUE), 
             data = decomp_data[decomp_data$treatment == "Control",])
pred_ctrl_k <- predict(ctrl_k, newdata = data.frame(time = pred_time))
lines(pred_time, pred_ctrl_k, col = "blue", lwd = 2)
# 拟合放牧样地二次模型
graz_k <- lm(k ~ poly(time, 2, raw = TRUE), 
             data = decomp_data[decomp_data$treatment == "Grazing",])
pred_graz_k <- predict(graz_k, newdata = data.frame(time = pred_time))
lines(pred_time, pred_graz_k, col = "red", lwd = 2)
legend("topright", legend = c("对照", "放牧"), 
       col = c("blue", "red"), pch = 16, lwd = 2)

# --- 指标Bg：二次多项式趋势 ---
plot(Bg ~ time, data = decomp_data, 
     col = ifelse(treatment == "Control", "blue", "red"),
     pch = 16, main = "Bg: 二次多项式趋势", 
     xlab = "时间", ylab = "Bg值")
ctrl_Bg <- lm(Bg ~ poly(time, 2, raw = TRUE), 
              data = decomp_data[decomp_data$treatment == "Control",])
pred_ctrl_Bg <- predict(ctrl_Bg, newdata = data.frame(time = pred_time))
lines(pred_time, pred_ctrl_Bg, col = "blue", lwd = 2)
graz_Bg <- lm(Bg ~ poly(time, 2, raw = TRUE), 
              data = decomp_data[decomp_data$treatment == "Grazing",])
pred_graz_Bg <- predict(graz_Bg, newdata = data.frame(time = pred_time))
lines(pred_time, pred_graz_Bg, col = "red", lwd = 2)
legend("topright", legend = c("对照", "放牧"), 
       col = c("blue", "red"), pch = 16, lwd = 2)

# --- 指标Br：二次多项式趋势 ---
plot(Br ~ time, data = decomp_data, 
     col = ifelse(treatment == "Control", "blue", "red"),
     pch = 16, main = "Br: 二次多项式趋势", 
     xlab = "时间", ylab = "Br值")
ctrl_Br <- lm(Br ~ poly(time, 2, raw = TRUE), 
              data = decomp_data[decomp_data$treatment == "Control",])
pred_ctrl_Br <- predict(ctrl_Br, newdata = data.frame(time = pred_time))
lines(pred_time, pred_ctrl_Br, col = "blue", lwd = 2)
graz_Br <- lm(Br ~ poly(time, 2, raw = TRUE), 
              data = decomp_data[decomp_data$treatment == "Grazing",])
pred_graz_Br <- predict(graz_Br, newdata = data.frame(time = pred_time))
lines(pred_time, pred_graz_Br, col = "red", lwd = 2)
legend("topright", legend = c("对照", "放牧"), 
       col = c("blue", "red"), pch = 16, lwd = 2)

关键说明 #

• poly(time, 2, raw=TRUE)：确保拟合的是常规二次多项式（time + time²），若不加raw=TRUE会生成正交多项式，不影响趋势线形状但系数解释逻辑不同。
• 指数模型初始参数：nls()需要初始值start，可根据你的数据初始S值设置a，衰减系数b设为0.1左右；若拟合报错，优先选择对数转换的线性模型方法。
• 连续预测序列：length.out=100保证趋势线足够平滑，可根据需求调整数值。

内容的提问来源于stack exchange，提问作者EBH