Changelog • CleanLASA

CleanLASA 1.0.0

First stable release.

transform_lasa_data() reshapes LASA data between wide format (one row per respondent, as in Z files) and long format (one row per respondent per wave). Long format has a Wave column and a Time column that counts waves from each respondent’s own cohort baseline (B, 2B, 3B, MB or 4B). Merged and stacked files work too.
read_lasa_sav(format_as_long = TRUE) returns a file in long format straight away.
Wide output can name columns by wave (prefix = "Wave") or by time since baseline (prefix = "Time"). Time-prefixed data carries a cohort column that merges with LASAZ002.sav’s.
Cross-wave Z files such as LASAZOA1.sav are now labelled and converted to factors.

The new keep_user_na argument decides what happens to missing-value codes, and it does the same thing for every variable type. With FALSE (the default), every missing code becomes NA. With TRUE, every missing code is kept, as label text or as a declared-missing haven_labelled_spss value. user_na is deprecated: a supplied value is used as keep_user_na, with a warning.
Discrete negative codes are always treated as missing, even when the label sounds substantive (e.g. “never had job” or “valid score”). Variables whose genuine values can be negative (e.g. dB thresholds) keep those values. The label database marks them with can_be_negative.

Fixed many gaps and errors in the bundled label database, including the ones reported in #18.
“na, see …” routing labels now name only the referenced variable.
Codes that have no wording are named after their number. Listed values of numeric variables that have no wording are dropped.
The bundled database is now found even without library(CleanLASA).