在R中编码NLS拟合温雨关系时遇维度匹配错误求助

Hey there! Let's walk through solving your nonlinear regression problems step by step—you're already on the right track with your logistic function, just a couple of small missteps in setting up nls().

First: Understand Your Model's Parameters #

Your func_rain(x,b) = 1/(1+exp(-b*x)) is a basic logistic curve, and it only has one parameter to estimate: b. The x (temperature) is your independent variable, and vec_rain/dp.w2$rain is your dependent variable—neither of these belong in the start argument! That's the root cause of both your errors.

Step 1: Fix the Vector-Based NLS Code #

Your first attempt threw a dimension mismatch error because you included vec_temp_num=2.6 in the start argument. Here's the corrected version:

# Define your function (unchanged)
func_rain <- function(x,b){(1/(1+exp(-b*x)))}

# Corrected vector-based fit
fit_vec <- nls(vec_rain ~ func_rain(vec_temp_num, b), 
               start = list(b = 1))  # Only include the parameter b!

Step 2: Fix the Data Frame-Based NLS Code #

Your second attempt had two issues: you added non-parameter variables (temperature, rain) to start, and you didn't use the data argument to simplify the formula. Here's the fixed code:

# Corrected data frame-based fit
Rain_fit <- nls(rain ~ func_rain(temperature, b), 
                data = dp.w2,  # Tell nls where to find your columns
                start = list(b = 1))  # Again, only parameter b

Optional (But Recommended): Add an Inflection Point Parameter #

Your initial attempt included 2.6—I suspect you wanted to set the inflection point (where rain proportion hits 50%) at 2.6 degrees. Your current function has its inflection point at x=0, which probably doesn't match your data. Let's adjust the function to include an inflection point parameter x0:

# Updated logistic function with inflection point
func_rain_inflect <- function(x, b, x0) {
  1 / (1 + exp(-b * (x - x0)))
}

# Fit with both parameters (b = slope, x0 = inflection temp)
Rain_fit_better <- nls(rain ~ func_rain_inflect(temperature, b, x0),
                       data = dp.w2,
                       start = list(b = 1, x0 = 2.6))  # Now we need both initial values

This will give you a model that's far more aligned with real-world temperature-precipitation relationships.

Step 3: Verify Your Fit #

Once you run the corrected code, check the results with:

summary(Rain_fit_better)  # See parameter estimates and model stats

You can also plot the fit to visualize how well it matches your data:

# Generate predicted values
dp.w2$pred_rain <- predict(Rain_fit_better, newdata = dp.w2)

# Plot raw data + fitted curve
plot(dp.w2$temperature, dp.w2$rain, 
     main = "Temperature vs Rain Proportion",
     xlab = "Temperature (°C)", ylab = "Rain Proportion")
lines(dp.w2$temperature, dp.w2$pred_rain, col = "darkred", lwd = 2)

Why Your Original Code Failed #

• First error: You included the independent variable (vec_temp_num) in start—nls() only expects parameters to estimate here, not data columns. This caused a dimension mismatch.
• Second error: You added both independent and dependent variables to start, and didn't use the data argument. This confused R's formula parser, leading to the "unexpected end of input" error.

内容的提问来源于stack exchange，提问作者Sanda