[Candice Yandam]

Jorrit:

Sure

* Example generated by -dataex-. To install: ssc install dataex
clear
input double hhcode long count int(street group)
31010101420086 310101 14 13
31010101420086 310101 14 13
31010101420086 310101 14 13
31010101420299 310101 14 13
31010101420299 310101 14 13
31010101420299 310101 14 13
31010101420354 310101 14 13
31010101420354 310101 14 13
31010101420354 310101 14 13
31010101420762 310101 14 13
31010101420762 310101 14 13
end
label values hhcode hhcode
label values count count
label values street street
label values group group

I'm don't know how to remove the value label of this variable. And I don't know what it means that a variable has a value label. This data is survey data.

[Jorrit Gosens]

Still a bit funny. The value label appears empty, at least on my side. In fact all of the value labels included with your sample do.
How were these value labels created?

What do you get when you type

label list

[Candice Yandam]

Jorrit:

I don't know how these value labels were created. I'm simply trying to use this survey data. This is what I get when I type

label list

label list
_merge:
           1 master only (1)
           2 using only (2)
           3 matched (3)
           4 missing updated (4)
           5 nonmissing conflict (5)
c11:
           1 Agriculture, Forestry, Animal husbandry
           2  Mining
           3 Manufacturing
           4 Production and Supply of Electricity, Gas and Water
           5 Construction Enterprise
           6 Transport, Storage and Post Industry
           7 Information Transmission, Computer Services and Software Industry
           8 Wholesale and Retail Trade
           9 Hotel and Catering Services
          10 Financial Intermediation
          11 Real Estate Industry

The list is very long but does not include the variable hhcode. Could it be possible that Stata believes that hhcode has a value label while it doesn't?

Thanks!

[Jorrit Gosens]

Well, more precise you could do

label list hhcode

If that returns ''value label hhcode not found" then its just an empty label here (i.e., no label values defined).

What i woudl do is.

ren hhcode hhcodeold
tostring hhcodeold, gen(hhcode)format("%14.0f")

This saves a copy of the original if you ever want to go back. Labels are not copied over when creating a new var like this.

[Nick Cox]

If hhcode was a previously encoded variable, then you have to go back to the original values.

[Jorrit Gosens]

I wondered about that being the origin of the label, but it didnt seem likely that this would create values in the range of 31010101420086.
Or is it?

[Nick Cox]

Jorrit: I agree; that sounds unlikely.

[Candice Yandam]

Jorrit:

Yes, hhcode had just an empty label.
I followed your suggestion and it worked! Many thanks!

Nick:

hhcode is encoded as "double" in the 2007 dataset but a string in the 2013 dataset.

[Candice Yandam]

Basically, part of the identifier refers to the county code, then street, and then individual id of the household which makes a 14 integers long number/identifier.

[Jorrit Gosens]

Alright.
Note that if you were entirely certain the label was not needed you could also do:

tostring hhcode, replace format("%14.0f") force

The suggestion to copy over the variable first is just a precaution against throwing away info from a dataset I am not familiar with and could only see a small part of.

[Candice Yandam]

Jorrit:

I understand, and I will keep this precaution just in case. Thanks!

Thank you all very much for your help