access label of value

[Florian Wakolbinger]

Dear all,
is there a way in stata to access the label of values in an if-condition?
for example, if I wanted to count the number of observations where var1 has a value whose value label is A?

something like
count if label(var1) == "A"

Thanks and best regards,
Florian

[Charlie Joyez]

Hello Florian and welcome to the list.

I don't know any direct way to do it, and I don't know if some exists, but the following code does what you want :

decode var1,gen(lab_var1)
count if lab_var=="A"

It creates a new Stata variable (which labels are not) corresponding to your label (hence, a string variable), and allows you to add the if exp when using it.

See a whole example here :

clear
set obs 5
gen num_var=_n

label define lbl 1 "one" 2 "two" 3"three" 4"four" 5"five"

label values num_var lbl
 
decode num,gen(lab_var)
count if lab_var=="one"

Hope this helps,
Charlie

[Joseph Coveney]

You can use the dereferencing syntax "<value>":<value label name>.

If you don't know the name of the set of value labels for a particular variable (or you want to do it programmatically), then you can couple that syntax with an extended macro function (help extended_fcn) that will find it for you.

Something like the following (at the "Begin here" comment).

ÿversionÿ14.2

.ÿ
.ÿclearÿ*

.ÿsetÿmoreÿoff

.ÿquietlyÿsetÿobsÿ4

.ÿgenerateÿbyteÿvar1ÿ=ÿmod(_n,ÿ2)

.ÿlabelÿdefineÿValuesÿ0ÿAÿ1ÿB

.ÿlabelÿvaluesÿvar1ÿValues

.ÿ
.ÿ*
.ÿ*ÿBeginÿhere
.ÿ*
.ÿcountÿifÿvar1ÿ==ÿ"A":`:ÿvalueÿlabelÿvar1'
ÿÿ2

.ÿ
.ÿexit

endÿofÿdo-file


.

[Sergiy Radyakin]

Joseph, in a perfect world that would be sufficient. But commonly you are dealing with 1 "small" 2 "small" 3 "small" etc, 97 "large" 98 "large" etc. Then Florian's question makes perfect sense.
Charlie's answer would handle this situation, but it requires creation of a temporary variable to hold the whole value label, which will not work well in older Stata's since value labels can be up to 32,000 long.

Best, Sergiy

[Clyde Schechter]

Tangential comment:

But commonly you are dealing with 1 "small" 2 "small" 3 "small" etc, 97 "large" 98 "large" etc

It's really interesting how we have very different experiences with data. In my entire career, I have never encountered a value label where the same label was used for different numeric codes, except when it turned out to be a mistake (and that usually only involved one label duplicated once.)

[Sergiy Radyakin]

Clyde, this stems from some data entry packages allowing application of labels to intervals, such as : 0-5 "Baby", 6-14 "Child", 15-60 "Adult", 61-125 "Senior". The resulting data set in Stata will have to have point labels to each of the integers from the interval (sadly Stata doesn't allow to label non-integers), resulting in massive number of duplicate labels.

I agree, creating a second variable agegroup with these 4 categories sounds more like Stata-World solution, but it is a second variable occupying more dataspace (if anyone still cares) and potentially out of sync with the original age variable in case that one gets edited (that is what I do care about).

Other popular choices are to provide skins:

• Labelset A: 101 beef, 102 pork, 103 veal, ... 201 shark, 202 tuna, ...
• Labelset B: 100-199 Meats, 200-299 Fish, ...

With subsequent tabulations based on the "active" labels.

"Commonly" is probably an exaggeration. I should have written "it is possible in general" or "it happens sometimes".

Best, Sergiy

[Florian Wakolbinger]

Thanks for your comments, very helpful. In my case, I don't have the problem that different integers have the same Label...
Best, Florian