如何解决nlsLM非线性回归时持续出现的起始值警告问题

问题原因 #

• 第一个警告Ignoring unknown parameters: start：geom_smooth本身不支持直接传入start参数，你写在geom_smooth里的start会被直接忽略。
• 第二个警告：因为nlsLM没有接收到指定的起始值，只能默认将a和b初始化为1，这才是你担心拟合偏差的根源。

解决方案 #

给nlsLM传递起始值必须通过geom_smooth的method.args参数，同时建议直接复用你之前手动拟合好的fitC5系数作为起始值，保证拟合结果和手动计算的一致。

修改后的完整代码 #

C5_abund <- read.csv(file = "C5_R.csv", sep=';')
C5_abund2 <- C5_abund[C5_abund$Day <= 9.82, ] #exponetial regresion applied on theses point

get_eqn <- function(fit) {
  eqn <- paste0("y = ", round(coef(fit)[1], 3), " * exp(", round(coef(fit)[2], 3), " * x)")
  return(eqn)
}


fitC5 <- nlsLM(C5 ~ a*exp(b*Day), data=C5_abund2, start=list(a=33804.52, b=0.5)) 
# using a= abundance at first day of experiment and b=theorical rate of growth
eqnC5 <- get_eqn(fitC5)
#y = 16470.994 * exp(0.525 * x)


rss <- sum(resid(fitC5)^2)
tss <- sum((C5_abund2$C5 - mean(C5_abund2$C5))^2)
rsq2C5 <- 1 - (rss/tss)

# rsq C5 :  0.9995


ggplot() +
  geom_point(data = C5_abund, aes(x = Day, y = C5),shape=16) +
  # 核心修改：将start参数放入method.args中传递给nlsLM
  geom_smooth(data = C5_abund2, aes(x = Day, y = C5), method = "nlsLM", 
              formula = y ~ a * exp(b * x), se = FALSE,
              method.args = list(start = list(a = coef(fitC5)[1], b = coef(fitC5)[2])),
              size=0.5, linetype = "dashed", color="black") +
  annotate("text", x = 2, y =800000, label = "atop(y == 16470.994 * e ^ {0.525*x}, R ^ 2 == 0.9995 )",
           parse = TRUE, color = "black", size =4) +
  labs(title = "Abundance Dynamic of C5 culture over days", x = "Time (days)", y = "Cell concentration (cell/mL)") +
  theme_base()+
  theme(legend.position = "bottom",
        plot.title = element_text(size = 14, hjust = 0.5),
        axis.title = element_text(size = 12),
        panel.grid.major = element_blank(), panel.grid.minor = element_blank(),
        legend.background = element_rect(fill = "white", size = 0.5, linetype = "solid",
                                         color = "black"))

效果验证 #

修改后运行代码，两个警告都会消失：

1. method.args会把起始值正确传递给nlsLM，避免默认初始化1的问题；
2. geom_smooth不再收到未知参数，也不会抛出忽略参数的警告。
   拟合出的曲线会和你手动计算的y = 16470.994 * exp(0.525 * x)完全一致，无需担心拟合准确性。

内容的提问来源于stack exchange，提问作者Marlexis