R语言中基于survey.design2/survey.design类加权调查数据绘制堆叠/分组条形图

Got it, let's work through this together! When dealing with survey.design or survey.design2 objects, regular ggplot2 syntax (like you'd use with raw data frames) doesn't work directly because ggplot can't interpret the weighted sampling design under the hood. The key fix is to first generate a weighted summary table with svytable(), then convert it into a tidy data frame that ggplot can handle seamlessly.

Step 1: Prepare Your Weighted Summary Table #

First, use svytable() to create your weighted cross-tabulation, then convert it to a standard data frame. Let's use a built-in example to make this concrete:

# Load required packages
library(survey)
library(ggplot2)
library(dplyr)

# Create a sample survey design (using the api dataset from the survey package)
data(api)
dclus1 <- svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc)

# Generate weighted cross-tabulation (here: school type vs. award eligibility)
svy_table <- svytable(~stype + awards, dclus1)

# Convert the svytable output to a regular data frame
tab_df <- as.data.frame(svy_table)

The resulting tab_df will have columns for your categorical variables (e.g., stype, awards) and a Freq column holding the weighted counts—this is the data structure ggplot2 needs.

Step 2: Build Grouped Bar Charts #

For side-by-side grouped bars, use position = position_dodge() to separate the bars for each category:

ggplot(tab_df, aes(x = stype, y = Freq, fill = awards)) +
  geom_bar(stat = "identity", position = position_dodge(width = 0.8)) +
  labs(
    title = "Weighted Grouped Bar Chart: School Type vs. Award Status",
    x = "School Type",
    y = "Weighted Frequency",
    fill = "Award Eligibility"
  ) +
  theme_minimal()

Step 3: Build Stacked Bar Charts #

For stacked bars, just switch to position = "stack" (this is ggplot's default, but writing it explicitly makes your intent clear):

ggplot(tab_df, aes(x = stype, y = Freq, fill = awards)) +
  geom_bar(stat = "identity", position = "stack") +
  labs(
    title = "Weighted Stacked Bar Chart: School Type vs. Award Status",
    x = "School Type",
    y = "Weighted Frequency",
    fill = "Award Eligibility"
  ) +
  theme_minimal()

Bonus: Add Percentages (Row/Column) #

If you want to plot percentages instead of raw weighted counts, calculate them first using dplyr:

# Calculate row percentages (percent of each award type within school type)
tab_df_pct <- tab_df %>%
  group_by(stype) %>%
  mutate(row_percent = (Freq / sum(Freq)) * 100)

# Plot grouped bar chart with row percentages
ggplot(tab_df_pct, aes(x = stype, y = row_percent, fill = awards)) +
  geom_bar(stat = "identity", position = position_dodge(width = 0.8)) +
  labs(
    title = "Weighted Grouped Bar Chart (Row Percentages)",
    x = "School Type",
    y = "Percentage",
    fill = "Award Eligibility"
  ) +
  theme_minimal()

Why This Works #

The core issue is that survey.design objects store metadata about the sampling design (weights, clusters, etc.) that ggplot2 doesn't natively understand. By using svytable(), we collapse the data into a weighted summary table that acts exactly like a regular data frame—perfect for working with ggplot2's grammar of graphics.

内容的提问来源于stack exchange，提问作者Flavia