The explanation is simple. "C" and "S" are not included in your -label-
definition. So, -encode- sees no labels of "C" and "S". So, it has to go
beyond your -label- definition.
You want -encode- to understand that "C" and "Control" mean the same as
far as you are concerned, and similarly "S" and "Salmon", but you
provide no information for Stata to know that. Otherwise put, Stata is
totally literal here in handling strings. "C" is not equal to Control.
What you want can be achieved with extra lines, e.g.
replace fish = "Control" if fish == "C"
replace fish = "Salmon" if fish == "S"
label def fish 0 "Control" 1 "Salmon"
encode fish, gen(fish1) label(fish)
Nick
[email protected]
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Miranda Kim
Sent: 24 July 2009 12:28
To: [email protected]
Subject: Re: st: RE: help with encode
Thank you for your input. As you say, there are many different ways to
make this work and so I have bypassed using encode.
I am grateful for the many suggestions, and I apologize if my initial
question was not laid out clearly enough.
Many thanks,
Miranda
ps: here is what I got trying to use encode:
. tab1 fish
-> tabulation of fish
fish | Freq. Percent Cum.
------------+-----------------------------------
C | 56 50.45 50.45
S | 55 49.55 100.00
------------+-----------------------------------
Total | 111 100.00
. desc fish
storage display value
variable name type format label variable label
------------------------------------------------------------------------
--------------
fish str1 %1s
. label def fish 0 "Control" 1 "Salmon"
. encode fish, gen(fish1) label(fish)
. tab1 fish1, nolabel
-> tabulation of fish1
fish1 | Freq. Percent Cum.
------------+-----------------------------------
2 | 56 50.45 50.45
3 | 55 49.55 100.00
------------+-----------------------------------
Total | 111 100.00
Michael Hanson wrote:
If that really is the situation you're in -- all encoded values are
2's and 3's -- then why not simply linearly transform the encoded
variables to the values that you want? That is, type:
replace gender = gender - 2
(Subtract 1 if you had 1's and 2's as mentioned in your original
post.)
However, I suspect something else is going on. My expectation is that
you have more than just two string values ("m" and "f") in your
series. Can you provide to the list the output of -table female-? I
suspect that if you try Nick's -tab ...- command shown below, you
would also find unexpected values.
However, these are all conjectures. You would likely resolve your
situation much more quickly (and waste fewer people's time on the list
in the process) if you followed the guidance clearly laid out in the
Statalist FAQ:
3.3 Stata references in your question
Say exactly what you typed and exactly what Stata typed (or did) in
response. N.B. exactly!
Why not at least show us what actually happened? In this case, copy
and paste from the results window the *exact* -encode- command and
resulting Stata output. Then copy and past whatever command led you
to conclude that you were getting 2's and 3's, and the *exact* and
*complete* output from that command. (In consideration to Statalist
readers, please don't provide us with a lengthy -list- output. Use
-table- instead, as suggested above.)
Hope this helps,
Mike
On Jul 24, 2009, at 4:09 AM, Miranda Kim wrote:
I tried this but couldn't make it work, as it then automatically
encodes the variables with 2's and 3's...
Nick Cox wrote:
In addition to other answers the direct answer to the second
question is
"Yes":
label def female 0 "m" 1 "f" encode gender, gen(female)
label(female)
It would no harm to check on any missings:
tab gender if !inlist(female, 0, 1)
Nick [email protected]
Miranda Kim
How can I efficiently convert string variables (such as gender with
values 'f' 'm') into binary 0/1 variables?
Can I fiddle with encode so that it codes 0/1 instead of 1/2?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
