如何导入变量定义在外部文件的固定宽度文件到R？YRBS数据导入求助

Hey there! Let's walk through how to import that YRBS fixed-width data into R, especially since we need to prep it for the survey package. This is a super common task with CDC datasets, so let's break it down step by step:

First, make sure you have two key files saved locally:

• The fixed-width data file (typically a .dat or .txt file from the YRBS page)
• The codebook/metadata file (this tells you variable names, start/end positions, value labels, and sampling details like weights, strata, and PSUs)

Fixed-width files depend on knowing exactly where each variable starts and ends. Instead of manually transcribing this, we can extract the specs from the codebook:

If your codebook is a text file: #

Use base R to load and parse it with regex (adjust the regex to match your codebook's formatting—YRBS codebooks usually have consistent lines with variable name, start, end):

# Load the codebook text
codebook_lines <- readLines("yrbs_codebook.txt")

# Extract lines with variable position info
var_spec_lines <- codebook_lines[grepl("^[A-Z0-9]+\\s+[0-9]+\\s+[0-9]+", codebook_lines)]

# Split each line into parts
var_specs <- strsplit(var_spec_lines, "\\s+")

# Pull out names, start positions, end positions
var_names <- sapply(var_specs, `[`, 1)
var_starts <- as.integer(sapply(var_specs, `[`, 2))
var_ends <- as.integer(sapply(var_specs, `[`, 3))

If your codebook is a PDF: #

Use the pdftools package to convert it to text first:

library(pdftools)

# Convert PDF to text
codebook_text <- pdf_text("yrbs_codebook.pdf")
codebook_lines <- unlist(strsplit(codebook_text, "\n"))

# Then follow the same regex extraction steps as above

With the variable specs ready, use read.fwf() to load the data correctly:

# Calculate column widths (end position - start position + 1)
var_widths <- var_ends - var_starts + 1

# Import the data
yrbs_raw <- read.fwf(
  file = "yrbs_data.dat",
  widths = var_widths,
  col.names = var_names,
  stringsAsFactors = FALSE,
  encoding = "ASCII" # YRBS files often use ASCII encoding
)

Since YRBS uses complex sampling, we need to create a survey design object with the survey package. First, confirm the weight, stratum, and PSU variable names from the codebook (they're usually something like WEIGHT, STRATUM, PSU):

library(survey)

# Create the survey design
yrbs_survey <- svydesign(
  id = ~PSU,
  strata = ~STRATUM,
  weights = ~WEIGHT,
  data = yrbs_raw,
  nest = TRUE # YRBS uses nested sampling, so this is important
)

YRBS variables use coded values (e.g., 1 = Yes, 2 = No). The labelled package makes it easy to add these labels for clarity:

library(labelled)

# Example: Add labels to a smoking variable
yrbs_raw$smoke <- set_value_labels(
  yrbs_raw$smoke,
  `1` = "Yes",
  `2` = "No",
  `9` = "Don't know"
)

• Column mismatch errors: Double-check your start/end positions—sometimes codebooks use 1-indexing, which is what read.fwf() expects, but if you're getting issues, confirm the positions again.
• Garbled text: If you see weird characters, make sure you're using the correct encoding (ASCII works for most YRBS files).
• Survey design errors: Ensure weight, stratum, and PSU variables are numeric. If they're stored as characters, convert them with as.integer() or as.numeric().

内容的提问来源于stack exchange，提问作者Chris