Skip to contents

Export Data to SPSS Format

Source: R/write_spss.R
write_spss.Rd

Writes a data frame to an SPSS .sav file, preserving variable labels, value labels, and user-defined missing values. When exporting data that was imported with read_spss() (with tag_na = TRUE), the tagged NAs are automatically converted back to SPSS user-defined missing values, enabling full roundtrip fidelity.

Usage

write_spss(data, path, compress = c("byte", "none", "zsav"))

Arguments

data

A data frame to export. Columns of class haven_labelled will have their labels and missing value metadata written to the .sav file.

path

Path to the output file. Must end in .sav or .zsav.

compress

Compression type. One of "byte" (default, byte-level compression), "none" (no compression), or "zsav" (zlib compression, requires SPSS v21+).

Value

Invisibly returns the file path.

Details

Tagged NA Roundtripping

Data imported via read_spss() stores SPSS user-defined missing values as tagged NAs with an na_tag_map attribute mapping tag characters to original codes (e.g., -9, -8). write_spss() reverses this process: tagged NAs are converted back to their original numeric codes, and the SPSS user-defined missing value specification is reconstructed so that the exported .sav file has the same missing value definitions as the original: read_spss() remembers the original definition (e.g. LOWEST THRU -1), which is written back as long as it still covers every missing code and no valid value. Otherwise up to 3 codes are written as discrete values; more codes use SPSS's "range plus one discrete value" form (the tightest that contains no valid value), announced in one message. If no such range exists, the export stops with an error instead of declaring valid values missing.

Cross-Format Export

When exporting data originally imported from Stata or SAS (with native extended missing values like .a-.z or .A-.Z), these cannot be represented as SPSS user-defined missing values. In this case, they are written as system missing (regular NA) with a warning.

Derived Variables

Arithmetic on an imported variable (e.g. x + 1 inside mutate()) keeps its tagged NAs but loses the code map, so the original codes are unknown. Such values are written as system missing with a warning naming the variables. rec() keeps the map; std(), center(), pomps() and to_numeric() return plain NA by design. Numeric columns that carry a bare "labels" attribute (without the haven_labelled class) are exported with their value labels.

When to Use This

Use write_spss() when you:

  • Need to export processed survey data back to SPSS format

  • Want to preserve user-defined missing value definitions

  • Need roundtrip fidelity: read_spss() -> processing -> write_spss()

See also

read_spss() for importing SPSS files, write_xlsx() for Excel export, write_stata() for Stata export, write_xpt() for SAS transport export, untag_na(), strip_tags()

Other data-export: write_stata(), write_xlsx(), write_xpt()

Examples

# \donttest{
if (requireNamespace("haven", quietly = TRUE)) {
  # Roundtrip: write to a temporary .sav, read back
  tmp <- tempfile(fileext = ".sav")
  write_spss(survey_data, tmp)
  data <- read_spss(tmp)

  # Export with zlib compression (smaller file, requires SPSS v21+)
  tmp_z <- tempfile(fileext = ".zsav")
  write_spss(survey_data, tmp_z, compress = "zsav")

  unlink(c(tmp, tmp_z))
}
#> ✔ Wrote 16 variables (2500 obs.) to file18f66101b5f.sav
#> ✔ Wrote 16 variables (2500 obs.) to file18f66697cd96.zsav
# }