This rang a bell. A while back the problem was described in very
similar
terms:
========================
Suppose you define value labels, attach them to variables, and then
later change your mind about attaching labels.
Even if you -drop- the value labels, according to Stata, for example
the
-describe- or -codebook- commands, the variables continue to have
labels
attached: it is just that those labels do not exist.
Labels can be detached from varname by
. label values varname
-labdtch- makes it easy to do this with a varlist.
-labdtch- does not drop value labels. For that, use -label drop-.
-labdtch- ignores any string variables in varlist.
Examples
--------
. labdtch foreign
The following detaches all value labels from variables (possibly
useful, but also dangerous):
. labdtch _all
=======================
-labdtch- is in a rag-bag of label utilities called -labutil-
on SSC.
. ssc desc labutil
. ssc inst labutil
Nick
[email protected]
Having said that, a -foreach- loop does the
same thing:
foreach v of var * {
cap label val `v'
}
Dimitriy V. Masterov
> This is not a Halloween pranl. I am actually having a
> problem with ghost
> value labels. I am not sure whence the problem arises, so I
> will start at
> the beginning:
>
> 1) I converted a dataset from SPSS to Stata
> 2) I dropped all of the value labels using "label drop
> _all". "label list"
> showed that there are no value labels.
> 3) I cleaned the data and labeled SOME of the values myself.
> 4) Some of the numeric variables that had a value label
> before I dropped
> them, still "have" them. For example, "codebook leved27" yields
>
> ------------------------------------------------------------
> ------------------------------------------------------------
> ------------------------------------------------------------
> --------------------
> leved27
> Highest Level of Education Completed (Self-Report)
> ------------------------------------------------------------
> ------------------------------------------------------------
> ------------------------------------------------------------
> --------------------
>
> type: numeric (byte)
> label: V3078, but label does not exist
>
> range: [1,99] units: 1
> unique values: 12 missing .: 6/123
>
> mean: 5.91453
> std. dev: 9.32976
>
> percentiles: 10% 25% 50%
> 75% 90%
> 2 2 3
> 9 10
> ------------------------------------------------------------
> ----------------------
>
>
>
> "des leved27" yields
>
> storage display value
> variable name type format label variable label
> ------------------------------------------------------------
> -------------------
> leved27 byte %9.0g V3078 Highest Level
> of Education
> Completed
> (Self-Report)
>
> However, V3078 does not exist, and not all variables are
> effected. When I
> attempt to drop the label by hand, Stata gives me an error
> message saying
> that it does not exist. In my analysis, I segregate the
> variables based on
> whether they have value labels (other then yesno) or not,
> and it would be
> helpful if I could do this. This also makes me worried that
> there are some
> deeper problems that are not merely inconvenient.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
