Coding suggestions

[Sanchez, Edgar]

Greetings once again!

 

 

            I appreciate the help the group has been giving me and I
wish to get some ideas on how to handle a particular data point.  As the
data comes from access the data element will most likely be numeric but
there is the possibility that the respondent used a string code.
Therefore I end up with a string variable and I am trying to think of a
way to parse through the data element to test if it is a number or
string. I would like to create a variable that has either the number
provided by the respondent or the string response recoded as a special
number representing their response. So anyone have any thoughts on how I
could approach this problem?

 

 

Edgar I Sanchez, Ph.D.

 

[hidden email]

http://www.apqceducation.org/

 

====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[ViAnn Beadle]

It's really easy to test for whether a stringvar contains a number or not.
Something like this will work:

compute numbervar=0.
if not(sysmis(number(stringvar,f8.2))) numbervar=1.

Note that you'd get a warning for every instance of stringvar not being a
number so you'd better increase your max warnings with the SET command.

But it's unclear to me what you want the result to be when a string is
encountered. Would you give us a more concrete example. What are you
expecting, how do you want to recode it, and what do you want to do with the
recoded response?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Sanchez, Edgar
Sent: Thursday, May 29, 2008 3:38 PM
To: [hidden email]
Subject: Coding suggestions

Greetings once again!





            I appreciate the help the group has been giving me and I
wish to get some ideas on how to handle a particular data point.  As the
data comes from access the data element will most likely be numeric but
there is the possibility that the respondent used a string code.
Therefore I end up with a string variable and I am trying to think of a
way to parse through the data element to test if it is a number or
string. I would like to create a variable that has either the number
provided by the respondent or the string response recoded as a special
number representing their response. So anyone have any thoughts on how I
could approach this problem?





Edgar I Sanchez, Ph.D.



[hidden email]

http://www.apqceducation.org/



=======
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Sanchez, Edgar]

Here is an example of what I am anticipating as far as responses go. I
have asked respondents how many hours are specifically dedicated toward
professional development in regard to utilization of some piece of
software. Some respondents will simply provide the number of hours they
have specifically set aside for this. For those who can not answer the
question because it is too difficult we ask them to respond by typing
T/D or N/A if they do not have that software in their organization. This
is an example I made up on the fly but it gives you an idea of what I am
anticipating. So can have responses like: "2 hrs" "2" "about 2 hrs"
"T/D" or "N/A". I don't have any data right now but I am trying to
prepare the code in anticipation of data analysis.

 

One of the hardest parts for me of transitioning from SAS to SPSS is
that the code for SPSS seems like disparate independent commands that
are not tied to datasets and it seems like independent request without a
logical flow. I know that this is largely due to the fact that I am just
learning SPSS code but perhaps it may be worth learning Python.

 

Edgar I Sanchez, PhD

 

[hidden email]

http://www.apqceducation.org/

 

 

-----Original Message-----
From: ViAnn Beadle [mailto:[hidden email]]
Sent: Thursday, May 29, 2008 5:54 PM
To: Sanchez, Edgar; [hidden email]
Subject: RE: Coding suggestions

 

It's really easy to test for whether a stringvar contains a number or
not.

Something like this will work:

 

compute numbervar=0.

if not(sysmis(number(stringvar,f8.2))) numbervar=1.

 

Note that you'd get a warning for every instance of stringvar not being
a

number so you'd better increase your max warnings with the SET command.

 

But it's unclear to me what you want the result to be when a string is

encountered. Would you give us a more concrete example. What are you

expecting, how do you want to recode it, and what do you want to do with
the

recoded response?

 

-----Original Message-----

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of

Sanchez, Edgar

Sent: Thursday, May 29, 2008 3:38 PM

To: [hidden email]

Subject: Coding suggestions

 

Greetings once again!

 

 

 

 

 

            I appreciate the help the group has been giving me and I

wish to get some ideas on how to handle a particular data point.  As the

data comes from access the data element will most likely be numeric but

there is the possibility that the respondent used a string code.

Therefore I end up with a string variable and I am trying to think of a

way to parse through the data element to test if it is a number or

string. I would like to create a variable that has either the number

provided by the respondent or the string response recoded as a special

number representing their response. So anyone have any thoughts on how I

could approach this problem?

 

 

 

 

 

Edgar I Sanchez, Ph.D.

 

 

 

[hidden email]

 

http://www.apqceducation.org/

 

 

 

=======

To manage your subscription to SPSSX-L, send a message to

[hidden email] (not to SPSSX-L), with no body text except the

command. To leave the list, send the command

SIGNOFF SPSSX-L

For a list of commands to manage subscriptions, send the command

INFO REFCARD

 

====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Richard Ristow]

At 06:53 PM 5/29/2008, ViAnn Beadle wrote:

>It's really easy to test for whether a stringvar contains a number or not.
>
>compute numbervar=0.
>if not(sysmis(number(stringvar,f8.2))) numbervar=1.
>
>Note that you'd get a warning for every instance of stringvar not
>being a number so you'd better increase your max warnings with the SET command.

Or decrease it, so you won't get all those warnings that tell you
things you already know.

Or, will this do it without the warnings? It's a quirk of RECODE that
I only just picked up reading the documentation, and I haven't tried it:

RECODE stringvar
    (CONVERT)
    (ELSE=SYSMIS) INTO numbervar.

That will give the correct numeric value, not 1, when the string is a
valid number. And I don't think it's supposed to give a warning.

-Onward with best wishes,
  Richard

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

[Art Kendall]

For a quick and dirty approach, start by doing something like this.
Look at the variable view  if a variable is string but you want to see
if it the kind of variable you are worried about
and it does not have legitimate negative values try something like this
untested syntax.

frequencies variables= odd, weird, messedup .
* eyeball the output and group strings with the same meaning into sets
on the recode statement.
recode odd('2 hrs', '2hr' =2)('1 day' =8)('T/D',
't/d'=-1)('N/A','n/a','na')('ABC','abc'=-3)('def'=-4)
(else=-5) (convert) into n_odd.
value labels
 -1 'T/D'
 -2 'N/A'
 -3 'abc'
 -4 'def'
 -5 'other missing'.
missing values n_odd (-999 thru -1).

OR
autorecode variables= odd, weird, messedup /into x y z.
recode x (...)(convert) into n_odd.

Art Kendall
Social Research Consultants

Sanchez, Edgar wrote:

> Here is an example of what I am anticipating as far as responses go. I
> have asked respondents how many hours are specifically dedicated toward
> professional development in regard to utilization of some piece of
> software. Some respondents will simply provide the number of hours they
> have specifically set aside for this. For those who can not answer the
> question because it is too difficult we ask them to respond by typing
> T/D or N/A if they do not have that software in their organization. This
> is an example I made up on the fly but it gives you an idea of what I am
> anticipating. So can have responses like: "2 hrs" "2" "about 2 hrs"
> "T/D" or "N/A". I don't have any data right now but I am trying to
> prepare the code in anticipation of data analysis.
>
>
>
> One of the hardest parts for me of transitioning from SAS to SPSS is
> that the code for SPSS seems like disparate independent commands that
> are not tied to datasets and it seems like independent request without a
> logical flow. I know that this is largely due to the fact that I am just
> learning SPSS code but perhaps it may be worth learning Python.
>
>
>
> Edgar I Sanchez, PhD
>
>
>
> [hidden email]
>
> http://www.apqceducation.org/
>
>
>
>
>
> -----Original Message-----
> From: ViAnn Beadle [mailto:[hidden email]]
> Sent: Thursday, May 29, 2008 5:54 PM
> To: Sanchez, Edgar; [hidden email]
> Subject: RE: Coding suggestions
>
>
>
> It's really easy to test for whether a stringvar contains a number or
> not.
>
> Something like this will work:
>
>
>
> compute numbervar=0.
>
> if not(sysmis(number(stringvar,f8.2))) numbervar=1.
>
>
>
> Note that you'd get a warning for every instance of stringvar not being
> a
>
> number so you'd better increase your max warnings with the SET command.
>
>
>
> But it's unclear to me what you want the result to be when a string is
>
> encountered. Would you give us a more concrete example. What are you
>
> expecting, how do you want to recode it, and what do you want to do with
> the
>
> recoded response?
>
>
>
> -----Original Message-----
>
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
>
> Sanchez, Edgar
>
> Sent: Thursday, May 29, 2008 3:38 PM
>
> To: [hidden email]
>
> Subject: Coding suggestions
>
>
>
> Greetings once again!
>
>
>
>
>
>
>
>
>
>
>
>             I appreciate the help the group has been giving me and I
>
> wish to get some ideas on how to handle a particular data point.  As the
>
> data comes from access the data element will most likely be numeric but
>
> there is the possibility that the respondent used a string code.
>
> Therefore I end up with a string variable and I am trying to think of a
>
> way to parse through the data element to test if it is a number or
>
> string. I would like to create a variable that has either the number
>
> provided by the respondent or the string response recoded as a special
>
> number representing their response. So anyone have any thoughts on how I
>
> could approach this problem?
>
>
>
>
>
>
>
>
>
>
>
> Edgar I Sanchez, Ph.D.
>
>
>
>
>
>
>
> [hidden email]
>
>
>
> http://www.apqceducation.org/
>
>
>
>
>
>
>
> =======
>
> To manage your subscription to SPSSX-L, send a message to
>
> [hidden email] (not to SPSSX-L), with no body text except the
>
> command. To leave the list, send the command
>
> SIGNOFF SPSSX-L
>
> For a list of commands to manage subscriptions, send the command
>
> INFO REFCARD
>
>
>
> ===================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD
>
>
>

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD