如何对鸢尾花数据集变量组合批量应用decisionplot并存储ggplot图

Hey there! Let's tackle that . not found error you're hitting with map2 and your decisionplot function. I’ve run into similar context mix-ups before, so let’s break this down step by step.

First, Let’s Diagnose the Root Cause #

The . not found error almost always happens when using map2’s formula shorthand (~) and trying to reference both input lists with the same . placeholder. R can’t tell which list you’re referring to, so it throws an error when it can’t resolve . correctly.

Let’s Fix the Code #

First, let’s make sure we’re on the same page with your setup. I’ll assume your decisionplot function looks something like this (adjust if yours differs):

library(ggplot2)

decisionplot <- function(model, data, xvar, yvar, ...) {
  # Create a grid of values for prediction
  x_grid <- seq(min(data[[xvar]]), max(data[[xvar]]), length.out = 100)
  y_grid <- seq(min(data[[yvar]]), max(data[[yvar]]), length.out = 100)
  prediction_grid <- expand.grid(x = x_grid, y = y_grid)
  colnames(prediction_grid) <- c(xvar, yvar)
  
  # Generate predictions
  prediction_grid$predicted <- predict(model, newdata = prediction_grid, type = "class")
  
  # Build the plot
  ggplot(data, aes(x = .data[[xvar]], y = .data[[yvar]], color = Species)) +
    geom_point(size = 2) +
    geom_tile(data = prediction_grid, aes(x = .data[[xvar]], y = .data[[yvar]], fill = predicted), 
              alpha = 0.2, inherit.aes = FALSE) +
    labs(x = xvar, y = yvar) +
    theme_minimal()
}

And your model/data list is built like this (handling Iris’s 3-class classification properly with nnet::multinom instead of binary glm):

library(tidyverse)
library(nnet)

# Get all non-repeating pairs of continuous variables
cont_vars <- iris %>% select(-Species) %>% colnames()
var_pairs <- combn(cont_vars, 2, simplify = FALSE) %>%
  set_names(map_chr(., ~paste(., collapse = " vs ")))

# Build list of (model, data) pairs for each variable combination
model_data_list <- map(var_pairs, function(pair) {
  data_subset <- iris %>% select(all_of(pair), Species)
  fitted_model <- multinom(Species ~ ., data = data_subset, trace = FALSE)
  list(data = data_subset, model = fitted_model)
})

Fix 1: Use ..1 and ..2 to Distinguish Inputs

In map2’s formula syntax, ..1 refers to elements from the first list (model_data_list), and ..2 refers to elements from the second list (var_pairs):

plots_list <- map2(model_data_list, var_pairs, 
                   ~decisionplot(model = ..1$model, 
                                 data = ..1$data, 
                                 xvar = ..2[1], 
                                 yvar = ..2[2]))

Fix 2: Use an Anonymous Function (More Readable)

If formula shorthand feels confusing, use a named anonymous function to explicitly define parameters—this eliminates ambiguity entirely:

plots_list <- map2(model_data_list, var_pairs, function(mdl_data, pair) {
  decisionplot(model = mdl_data$model,
               data = mdl_data$data,
               xvar = pair[1],
               yvar = pair[2])
})

Display the Plots in a Grid #

Use the patchwork package to arrange all 6 plots neatly:

library(patchwork)
wrap_plots(plots_list, ncol = 2) # Adjust ncol to change grid layout

Key Takeaway #

The . placeholder only works for single-input maps (map()). For map2(), you need to either use positional placeholders (..1, ..2) or explicitly name your function parameters to avoid context confusion.

内容的提问来源于stack exchange，提问作者user113156