10:02 a.m.
Dears,
If I have a string variable with the following mapping:
1 = 'middle'
2 = 'low'
3 = 'high'
there is a command to give the series a new, more logical mapping? Like as:
1 = 'low'
2 = 'middle'
3 = 'high'
I've simplified the example, but the concept is: changing the numerical
mapping of the entire series
Paolo
10:18 a.m.
Hi Paolo,
here is an example on how to do the mapping:
<hansl>
set verbose off
clear
nulldata 50
set seed 1234
series y = ceil(randgen(u, 0, 3)) # sample series with 3
disctinct integer values
strings label = defarray("a", "b", "c") # mapping: 1=a, 2=b,
3=c
stringify(y, label) # do the mapping
print y -o
/* Output
y
1 c
2 b
3 b
4 c
5 b
*/
</hansl>
I guess you need to change the numbering of "y" and the string array
"label", respectively.
This may be done as follows:
<hansl>
# swap categorie 2 and 3
matrix my = {y}
series y_reordered = replace(y, {2, 3}, {3,2})
stringify(y_reordered, label)
print y y_reordered -o
/* Output
y y_reordered
1 c b
2 b c
3 b c
4 c b
5 b c
*/
</hansl>
Artur
Am 01.02.2023 11:02 schrieb Paolo Chirico:
> Dears,
> If I have a string variable with the following mapping:
> 1 = 'middle'
> 2 = 'low'
> 3 = 'high'
> there is a command to give the series a new, more logical mapping? Like
> as:
> 1 = 'low'
> 2 = 'middle'
> 3 = 'high'
> I've simplified the example, but the concept is: changing the
> numerical mapping of the entire series
>
> Paolo
> _______________________________________________
> Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
> To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
> Website:
> https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
2:03 p.m.
On Wed, Feb 1, 2023 at 5:02 AM Paolo Chirico <paolo.chirico(a)uniupo.it> wrote:
Dears,
If I have a string variable with the following mapping:
1 = 'middle'
2 = 'low'
3 = 'high'
there is a command to give the series a new, more logical mapping? Like as:
1 = 'low'
2 = 'middle'
3 = 'high'
I've simplified the example, but the concept is: changing the numerical
mapping of the entire series
Here's my take on this. Suppose we have a file strs.csv with this content:
<csv>
x
middle
low
high
high
low
middle
</csv>
Then we can process it in this way:
<hansl>
open strs.csv
series numx = x
series y = x
# swap numerical values 1 and 2
series y = y==1 ? 2 : y==2 ? 1 : y
# add strings in 'logical' order
stringify(y, defarray("low", "middle", "high"))
series numy = y
print -o
</hansl>
(The "num*" series are added just to expose the numeric codings.)
Allin
3:12 p.m.
On Wed, Feb 1, 2023 at 9:03 AM Cottrell, Allin <cottrell(a)wfu.edu> wrote:
On Wed, Feb 1, 2023 at 5:02 AM Paolo Chirico <paolo.chirico(a)uniupo.it> wrote:
>
> Dears,
> If I have a string variable with the following mapping:
> 1 = 'middle'
> 2 = 'low'
> 3 = 'high'
> there is a command to give the series a new, more logical mapping? Like as:
> 1 = 'low'
> 2 = 'middle'
> 3 = 'high'
> I've simplified the example, but the concept is: changing the numerical
> mapping of the entire series
[...]
While my previous answer works for the simplified case, it may be
worth giving a more general, automated solution. Here's a function
that does the job:
<hansl>
function void recode (series *y, series x, strings S)
y = 0
matrix m = {y}
strings S0 = x
loop i=1..nelem(S)
matrix sel = instrings(S0, S[i])
m[sel] = i
endloop
y = m
stringify(y, S)
end function
</hansl>
The idea is that you pass (1) a target series in pointer form, (2) the
original string-valued series and (3) an array holding the string
values of the original series, re-ordered as per the desired numerical
coding. So in relation to my previous example you'd do:
<caller>
open strs.csv
series y
recode(&y, x, defarray("low", "middle", "high"))
</caller>
Allin
8:45 a.m.
If I understand correctly, you can't recode the same variable, but you
can create a new variable with the same categories and a different mapping.
Thanks for the tips
Paolo
Il 01/02/2023 16:12, Cottrell, Allin ha scritto:
On Wed, Feb 1, 2023 at 9:03 AM Cottrell, Allin
<cottrell(a)wfu.edu> wrote:
> On Wed, Feb 1, 2023 at 5:02 AM Paolo Chirico <paolo.chirico(a)uniupo.it> wrote:
>> Dears,
>> If I have a string variable with the following mapping:
>> 1 = 'middle'
>> 2 = 'low'
>> 3 = 'high'
>> there is a command to give the series a new, more logical mapping? Like as:
>> 1 = 'low'
>> 2 = 'middle'
>> 3 = 'high'
>> I've simplified the example, but the concept is: changing the numerical
>> mapping of the entire series
[...]
While my previous answer works for the simplified case, it may be
worth giving a more general, automated solution. Here's a function
that does the job:
<hansl>
function void recode (series *y, series x, strings S)
y = 0
matrix m = {y}
strings S0 = x
loop i=1..nelem(S)
matrix sel = instrings(S0, S[i])
m[sel] = i
endloop
y = m
stringify(y, S)
end function
</hansl>
The idea is that you pass (1) a target series in pointer form, (2) the
original string-valued series and (3) an array holding the string
values of the original series, re-ordered as per the desired numerical
coding. So in relation to my previous example you'd do:
<caller>
open strs.csv
series y
recode(&y, x, defarray("low", "middle", "high"))
</caller>
Allin
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
--
Paolo Chirico
Università del Piemonte Orientale
Dip. di Giurisprudenza e Scienze Politiche,
Economiche e Sociali (DIGSPES)
Alessandria, Italia
11:32 a.m.
Am 02.02.2023 um 09:45 schrieb Paolo Chirico:
If I understand correctly, you can't recode the same variable,
but you
can create a new variable with the same categories and a different
mapping.
Well, if you start out with series y and then have the recoded result in
x, you can always copy x over into y, and then you have y recoded.
cheers
sven
1:19 p.m.
On Thu, Feb 2, 2023 at 6:32 AM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
Am 02.02.2023 um 09:45 schrieb Paolo Chirico:
If I understand correctly, you can't recode the same variable, but you can create a
new variable with the same categories and a different mapping.
Well, if you start out with series y and then have the recoded result in x, you can
always copy x over into y, and then you have y recoded.
Actually, no. There's a rule against overwriting an entire
string-valued series. Maybe that could be relaxed, but also it might
be nice to have a recode() function that would achieve that with
proper error-checking (the target series won't get broken). The
arguments would be a pointer-to-series and an array of strings in the
desired order.
Allin
2:01 p.m.
Yes, you can always copy the new series over the old after the recoding
Paolo
Il giorno gio 2 feb 2023 alle 14:19 Cottrell, Allin <cottrell(a)wfu.edu> ha
scritto:
On Thu, Feb 2, 2023 at 6:32 AM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
>
> Am 02.02.2023 um 09:45 schrieb Paolo Chirico:
>
> If I understand correctly, you can't recode the same variable, but you
can create a new variable with the same categories and a different mapping.
>
> Well, if you start out with series y and then have the recoded result in
x, you can always copy x over into y, and then you have y recoded.
Actually, no. There's a rule against overwriting an entire
string-valued series. Maybe that could be relaxed, but also it might
be nice to have a recode() function that would achieve that with
proper error-checking (the target series won't get broken). The
arguments would be a pointer-to-series and an array of strings in the
desired order.
Allin
_______________________________________________
Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/
1016
days inactive
1017
days old
7 comments
4 participants
participants (4)
-
atecon -
Cottrell, Allin -
Paolo Chirico -
Sven Schreiber