SPSSX Discussion

One question about multiple response

Classic

List

Threaded

37 messages Options

Marta García-Granero

Re: Differences in variable values

Hi Sibusiyo

Tuesday, August 29, 2006, 11:13:00 PM, You wrote:

SM> Well I have two series of numeric data which I was hoping are
SM> similar/highly correlated/the same so I could use them
SM> interchangeably in my analysis.

First of all, high correlation doesn't mean "similar" or "the same".
The second series values might duplicate the first series (clearly,
quite a difference) but still the correlation might be very high. You
can use Lin's CCC (Concordance Correlation Coefficient), a type of
correlation coefficient that penalizes the departure of the series
from the equality line (I have code for it, or you can use
"Pairsetc.exe", part of WinPepi, a freeware statistical package:
http://www.brixtonhealth.com/pepi4windows.html ).

You could also try Bland-Altman analysis. Basically, you plot the
differences (y axis) against the average of the two values (x axis)
and take a look at the patterns (for instance: is the difference
independent of the mean values or is it correlated in some way?). You
can get a lot of information from the graph. Again, I have some code
(or you can try a demo of MedCalc http://www.medcalc.be/download.php )

--
Regards,
Dr. Marta García-Granero,PhD mailto:[hidden email]
Statistician

---
"It is unwise to use a statistical procedure whose use one does
not understand. SPSS syntax guide cannot supply this knowledge, and it
is certainly no substitute for the basic understanding of statistics
and statistical thinking that is essential for the wise choice of
methods and the correct interpretation of their results".

(Adapted from WinPepi manual - I'm sure Joe Abrahmson will not mind)

Sibusiso Moyo

Re: Creating a date variable

In reply to this post by Sibusiso Moyo

Dear ALL,

I have 10 products that I have been tracking for the past 60 months. So for each product I want to create a date variable that counts month 1 to month 60. I then want this repeated for products 2 to 10.

I'd like to creat a date variable of the following format.

I have a start date 08/01/2001 for case 1. Now for cases 2 up to 60 I want the date variable to add an additional month, so I could move from 08/01/2001 to 09/01/2001 etc until the 60th month which is 07/01/2006.

Adios,

Sibusiso.

Beadle, ViAnn

Re: Creating a date variable

see if the date command will do what you want

________________________________

From: SPSSX(r) Discussion on behalf of Sibusiso Moyo
Sent: Wed 8/30/2006 9:32 AM
To: [hidden email]
Subject: Re: Creating a date variable

Dear ALL,

I have 10 products that I have been tracking for the past 60 months. So for each product I want to create a date variable that counts month 1 to month 60. I then want this repeated for products 2 to 10.

I'd like to creat a date variable of the following format.

I have a start date 08/01/2001 for case 1. Now for cases 2 up to 60 I want the date variable to add an additional month, so I could move from 08/01/2001 to 09/01/2001 etc until the 60th month which is 07/01/2006.

Adios,

Sibusiso.

Maguin, Eugene

Re: Creating a date variable

In reply to this post by Sibusiso Moyo

Subusiso,

Here is one way to do this. The idea is to increment the months and years in
separate variables and then put them together. Days looks to be fixed to a
value of 1. So that is a constant. I'll assume your cases are sorted in the
correct order so that the case number is the month counter.

Compute mo1a=8+$casenum-1. /* this your starting month.
Compute mo1b=mod((mo1a-1),12)+1.
Compute yr1a=2001+trunc((mo1a-1)/12).
Compute date1=date.mdy(mo1b,1,yr1a).

Since you have multiple ones of these to do, you could use a Do repeat
structure. The key element in that setup is that numbers and characters as
well as variables can be used in the list of elements.

Gene Maguin

Sibusiso Moyo

Re: Creating a date variable

In reply to this post by Sibusiso Moyo

Thanks to ViAnn, I turned to the DATE Function in SPSS and it worked. I split my file by ID1 then allowed the date function to do its magic. Thanks again!

*** Split file
SORT CASES BY ID1 .
SPLIT FILE
SEPARATE BY ID1 .

The following new variables are being created:

Name Label

YEAR_ YEAR, not periodic
MONTH_ MONTH, period 12
DATE_ Date. Format: "MMM YYYY"
Rename variables DATE_ = MONTH.
EXE.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Gene Maguin
Sent: Wednesday, August 30, 2006 11:02 AM
To: [hidden email]
Subject: Re: Creating a date variable

Subusiso,

Here is one way to do this. The idea is to increment the months and years in
separate variables and then put them together. Days looks to be fixed to a
value of 1. So that is a constant. I'll assume your cases are sorted in the
correct order so that the case number is the month counter.

Compute mo1a=8+$casenum-1. /* this your starting month.
Compute mo1b=mod((mo1a-1),12)+1.
Compute yr1a=2001+trunc((mo1a-1)/12).
Compute date1=date.mdy(mo1b,1,yr1a).

Since you have multiple ones of these to do, you could use a Do repeat
structure. The key element in that setup is that numbers and characters as
well as variables can be used in the list of elements.

Gene Maguin

Sibusiso Moyo

Re: Learning to program in SPSS macros

In reply to this post by Sibusiso Moyo

Dear All,

I use SPSS 14 everyday, and I must admit that I am tired of writting un-intelligent codes. So I am trying to learn programming in SPSS macros. Any suggestions in terms of what material I should look at for examples for a beginner?

I currently have the 2003 version of SPSS Programming and Data Management by Ray Levesque.

Sibusiso.

Marta García-Granero

Re: Learning to program in SPSS macros

Hi Sibusiso

SM> I use SPSS 14 everyday, and I must admit that I am tired of
SM> writting un-intelligent codes. So I am trying to learn programming
SM> in SPSS macros.

Wellcome to the club.

SM> Any suggestions in terms of what material I should
SM> look at for examples for a beginner?

The Command Syntax Reference is quite useful, at least, it's what I
used when I started programming macros. The DEFINE-!ENDDEFINE chapter
has a lot of examples, starting with very simple MACROs. There is also
another Chapter at the end: Using the MACRO facility with 3 examples
explained step by step.

Take also a look at the MACRO section of Ray's website:

http://www.spsstools.net/Macros.htm

SM> I currently have the 2003 version of SPSS Programming and
SM> Data Management by Ray Levesque.

Since you are using SPSS 14, if you want to learn how to use the
Programmability Extension, you should download the 3rd Edition of the
book

http://www.spss.com/spss/data_management_book.htm

If you want to learn, for the moment, only "classic" macro
programming, then the 2003 edition of the book is OK.

--
Regards,
Marta

Sibusiso Moyo

Re: Learning to program in SPSS macros

In reply to this post by Sibusiso Moyo

Once again your advice is appreciated,

Thanks!

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Marta García-Granero
Sent: Thursday, August 31, 2006 5:42 AM
To: [hidden email]
Subject: Re: Learning to program in SPSS macros

Hi Sibusiso

SM> I use SPSS 14 everyday, and I must admit that I am tired of
SM> writting un-intelligent codes. So I am trying to learn programming
SM> in SPSS macros.

Wellcome to the club.

SM> Any suggestions in terms of what material I should
SM> look at for examples for a beginner?

The Command Syntax Reference is quite useful, at least, it's what I
used when I started programming macros. The DEFINE-!ENDDEFINE chapter
has a lot of examples, starting with very simple MACROs. There is also
another Chapter at the end: Using the MACRO facility with 3 examples
explained step by step.

Take also a look at the MACRO section of Ray's website:

http://www.spsstools.net/Macros.htm

SM> I currently have the 2003 version of SPSS Programming and
SM> Data Management by Ray Levesque.

Since you are using SPSS 14, if you want to learn how to use the
Programmability Extension, you should download the 3rd Edition of the
book

http://www.spss.com/spss/data_management_book.htm

If you want to learn, for the moment, only "classic" macro
programming, then the 2003 edition of the book is OK.

--
Regards,
Marta

Keith McCormick

Re: Learning to program in SPSS macros

Hello all,

I agree with all of the advice given already. I would simply add to
that the advice that you can visit, http://www.spss.com/devcentral/

This would be a good place to visit to determine if the "classic"
approach would suffice. Certain aspects of macros will be trumped by
the newer choices offered in 14.0., referred to as "programmability"
(discussed only in the 3rd edition). That vague term is used because
as of 14.0 external programming languages can be used, but there are
choices as to which one. As of the summer of this year, the two
choices are Python and VB.net. There hasn't been tremendous traffic
on the list serve regarding this yet, but it is bound to pick up since
it has been announced that 15.0 will have python on the SPSS
installation disk (or you can visit www.python.org and download it for
free). There are also "scripts" to consider.

Investing a couple of hours understanding you options, would be well
worth it, although the wealth of choices can seem a mixed blessing.

If time permits, consider the SPSS course that overviews all of the
programming options including macros, scripts, and python.

http://www.spss.com/training/pub_desc.cfm?courseID=100000436

I should note that I am aware of the course because I have taught it,
but I consider it a good investment of two days.

Good luck.

best, Keith

www.keithmccormick.com

On 8/31/06, Sibusiso Moyo <[hidden email]> wrote:

> Once again your advice is appreciated,
>
> Thanks!
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
> Marta García-Granero
> Sent: Thursday, August 31, 2006 5:42 AM
> To: [hidden email]
> Subject: Re: Learning to program in SPSS macros
>
>
> Hi Sibusiso
>
> SM> I use SPSS 14 everyday, and I must admit that I am tired of
> SM> writting un-intelligent codes. So I am trying to learn programming
> SM> in SPSS macros.
>
> Wellcome to the club.
>
> SM> Any suggestions in terms of what material I should
> SM> look at for examples for a beginner?
>
> The Command Syntax Reference is quite useful, at least, it's what I
> used when I started programming macros. The DEFINE-!ENDDEFINE chapter
> has a lot of examples, starting with very simple MACROs. There is also
> another Chapter at the end: Using the MACRO facility with 3 examples
> explained step by step.
>
> Take also a look at the MACRO section of Ray's website:
>
> http://www.spsstools.net/Macros.htm
>
> SM> I currently have the 2003 version of SPSS Programming and
> SM> Data Management by Ray Levesque.
>
> Since you are using SPSS 14, if you want to learn how to use the
> Programmability Extension, you should download the 3rd Edition of the
> book
>
> http://www.spss.com/spss/data_management_book.htm
>
> If you want to learn, for the moment, only "classic" macro
> programming, then the 2003 edition of the book is OK.
>
>
> --
> Regards,
> Marta
>

Keith McCormick

(no subject)

In reply to this post by Sibusiso Moyo

Hi Sibusiso,

Have you considered using the "Visual Bander" on the difference?

You could take the difference and explore which cases were outside +-
3 sd on the difference.

The standard deviation is easy to calculate using Analyze ...
Descriptive Statistics ..Descriptives.

The Visual Bander is found under Transform ... Visual Binning ....
You can make cutpoints at +- 3z (you don't need descriptive to do
this), and thereby divide your differences variable into three groups:
quite low, medium, and quite high.

Hope that helps,

Keith

www.keithmccormick.com

On 8/29/06, Sibusiso Moyo <[hidden email]> wrote:
> Dear All,
>
> I have about 13000 cases in my data set. I have these two variables that I am trying to test how close they are from each other. Specifically I am trying to identify cases, where the biggest difference arises. Besides taking differences between the two, sorting the data by the resulting difference variable, plots etc, would anyone have any ideas how one would proceed with such an assignment.
>
> Thanks,
>
> Sibusiso.
>

Sibusiso Moyo

Re: Learning to program in SPSS macros

In reply to this post by Sibusiso Moyo

Thanks for information Keith. I don't consider myself a programmer (am not proficient enough), although I must admit that I find programming very attractive as a past time. So besides learning the macros, I will be trying the Python extension. It looks very useful.

Sibusiso.

-----Original Message-----
From: Keith McCormick [mailto:[hidden email]]
Sent: Monday, September 04, 2006 12:57 PM
To: Sibusiso Moyo
Cc: [hidden email]
Subject: Re: Learning to program in SPSS macros

Hello all,

I agree with all of the advice given already. I would simply add to
that the advice that you can visit, http://www.spss.com/devcentral/

This would be a good place to visit to determine if the "classic"
approach would suffice. Certain aspects of macros will be trumped by
the newer choices offered in 14.0., referred to as "programmability"
(discussed only in the 3rd edition). That vague term is used because
as of 14.0 external programming languages can be used, but there are
choices as to which one. As of the summer of this year, the two
choices are Python and VB.net. There hasn't been tremendous traffic
on the list serve regarding this yet, but it is bound to pick up since
it has been announced that 15.0 will have python on the SPSS
installation disk (or you can visit www.python.org and download it for
free). There are also "scripts" to consider.

Investing a couple of hours understanding you options, would be well
worth it, although the wealth of choices can seem a mixed blessing.

If time permits, consider the SPSS course that overviews all of the
programming options including macros, scripts, and python.

http://www.spss.com/training/pub_desc.cfm?courseID=100000436

I should note that I am aware of the course because I have taught it,
but I consider it a good investment of two days.

Good luck.

best, Keith

www.keithmccormick.com

On 8/31/06, Sibusiso Moyo <[hidden email]> wrote:

Sibusiso Moyo

Re: Learning to program in SPSS macros

In reply to this post by Sibusiso Moyo

Keith,

As a follow up question. How compatible is Python to SPSS? Can all the stuff that I do on SPSS be transferred to Python? Which is one more user friendly (in my company people like things simple :))? Can the Python code be easity re-integrated to SPSS syntax?

Sibusiso.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Keith McCormick
Sent: Monday, September 04, 2006 12:57 PM
To: [hidden email]
Subject: Re: Learning to program in SPSS macros

Hello all,

I agree with all of the advice given already. I would simply add to
that the advice that you can visit, http://www.spss.com/devcentral/

This would be a good place to visit to determine if the "classic"
approach would suffice. Certain aspects of macros will be trumped by
the newer choices offered in 14.0., referred to as "programmability"
(discussed only in the 3rd edition). That vague term is used because
as of 14.0 external programming languages can be used, but there are
choices as to which one. As of the summer of this year, the two
choices are Python and VB.net. There hasn't been tremendous traffic
on the list serve regarding this yet, but it is bound to pick up since
it has been announced that 15.0 will have python on the SPSS
installation disk (or you can visit www.python.org and download it for
free). There are also "scripts" to consider.

Investing a couple of hours understanding you options, would be well
worth it, although the wealth of choices can seem a mixed blessing.

If time permits, consider the SPSS course that overviews all of the
programming options including macros, scripts, and python.

http://www.spss.com/training/pub_desc.cfm?courseID=100000436

I should note that I am aware of the course because I have taught it,
but I consider it a good investment of two days.

Good luck.

best, Keith

www.keithmccormick.com

On 8/31/06, Sibusiso Moyo <[hidden email]> wrote:

Richard Ristow

Re: Learning to program in SPSS macros

At 05:44 PM 9/5/2006, Sibusiso Moyo wrote:

>As a follow up question. How compatible is Python to SPSS? Can all the
>stuff that I do on SPSS be transferred to Python? Which is one more
>user friendly (in my company people like things simple :))? Can the
>Python code be easity re-integrated to SPSS syntax?

The short answer is, they're entirely different languages; in that
sense, not compatible at all. Nothing written to run in SPSS can run in
Python, or vice versa.

Python can create and submit SPSS statements that are then executed.
For how that's done; and for how different the syntaxes look; see my
just-now posting "Re: Renaming variables - noncontinuous", or various
postings by Jon Peck of SPSS, Inc.

It's being suggested that SPSS syntax be replaced with Python entirely;
that is, SPSS code will only be generated by Python programs, which
will use their own functions so that the users see none of the SPSS
code. For advocacy of that, see Jon Peck; for myself, I'm doubtful, for
anything like the near future. Among other things, it means a very
careful set of Python functions; and it means discarding much of the
skill of current SPSS programmers.

For more details, from a 'pro' point of view, see
http://www.spss.com/devcentral/files/articles/convert_spss_to_python1.pdf

Peck, Jon

Re: Learning to program in SPSS macros

It is true that Python and SPSS syntax are different languages, but it is very powerful to use Python to generate SPSS commands as well as using the direct functionality of Python function calls to SPSS.

But I'd like to clarify the "It's being suggested comment below". There is no intention of replacing SPSS syntax with Python syntax entirely if that is taken to mean that SPSS syntax would no longer be available. What is interesting is the possibility of providing a pure Python set of objects that would allow a Python programmer to write object-oriented code that would generate and submit SPSS syntax underneath.

For example, you can open a data file and assign a dataset name to it with SPSS syntax like this.

GET FILE='filespec'.
DATASET NAME mydataset.

Nothing wrong with that. But you could also accomplish this in a BEGIN PROGRAM block with

spssaux.OpenDataFile('filespec', dataset='mydataset')

In similar ways you can wrap more complex SPSS syntax in Python notation. The Transform module on Developer Central (www.spss.com/devcentral) allows you to do COMPUTE and DO IF commands along with VARIABLE LABEL, VARIABLE LEVEL, etc in o-o notation with something like this.

This Python syntax defines all the metadata for a new variable, computes it conditionally, and as an extra, makes the data updatable in the future.

newvar = Transform.Compute(varname="average_increase",
varlabel="Salary increase per month of experience if at least a year",\
varmeaslvl="Scale",\
varmissval=[999,998,997],\
varformat="F8.4")
newvar.expression = "(salary-salbegin)/jobtime"
newvar.condition = "jobtime > 12"
newvar.retransformable=True
newvar.generate() # An exception will be raised if the compute fails

The equivalent SPSS syntax (without the updatable variable) would be this.

NUMERIC average_increase.
FORMAT average_increase (F8.4).
VARIABLE LABEL average_increase "Salary increase per month of experience if at least a year".
VARIABLE LEVEL average_increase (Scale).
MISSING VALUES average_increase (999,998, 997).
DO IF jobtime > 12.
COMPUTE average_increase = (salary-salbegin)/jobtime.
END IF.

So this approach turns SPSS syntax into a full fledged object-oriented programming language, but it leaves SPSS syntax underneath doing the work.

And it certainly does not eliminate the powerful SPSS syntax, which will remain fully available. Not only is it widely known, there are zillions of existing jobs that will continue to run. And new SPSS syntax will continue to be created for new procedures. But if your style inclinations are towards object-oriented programming, the Python approach may appeal strongly. Someone, of course, would have to create the Python code to generate all that SPSS syntax, which is not happening now, at least within SPSS.

So this is an idea that I like, but that's my personal opinion.

Regards,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Richard Ristow
Sent: Wednesday, September 06, 2006 12:01 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Learning to program in SPSS macros

At 05:44 PM 9/5/2006, Sibusiso Moyo wrote:

>As a follow up question. How compatible is Python to SPSS? Can all the
>stuff that I do on SPSS be transferred to Python? Which is one more
>user friendly (in my company people like things simple :))? Can the
>Python code be easity re-integrated to SPSS syntax?

The short answer is, they're entirely different languages; in that
sense, not compatible at all. Nothing written to run in SPSS can run in
Python, or vice versa.

Python can create and submit SPSS statements that are then executed.
For how that's done; and for how different the syntaxes look; see my
just-now posting "Re: Renaming variables - noncontinuous", or various
postings by Jon Peck of SPSS, Inc.

It's being suggested that SPSS syntax be replaced with Python entirely;
that is, SPSS code will only be generated by Python programs, which
will use their own functions so that the users see none of the SPSS
code. For advocacy of that, see Jon Peck; for myself, I'm doubtful, for
anything like the near future. Among other things, it means a very
careful set of Python functions; and it means discarding much of the
skill of current SPSS programmers.

For more details, from a 'pro' point of view, see
http://www.spss.com/devcentral/files/articles/convert_spss_to_python1.pdf

Bauer, Craig

Re: Learning to program in SPSS macros

In reply to this post by Sibusiso Moyo

Jon said:
"Someone, of course, would have to create the Python code to generate
all that SPSS syntax, which is not happening now, at least within SPSS."

Can you please tell us more about this? What % of transformation and
basic proc commands are currently mapped to python methods/properties
(e.g. could someone abandon SPSS code entirely now for simple things)?
Will SPSS be completing this mapping for future releases?

Thanks

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Peck, Jon
Sent: Wednesday, September 06, 2006 8:02 AM
To: [hidden email]
Subject: Re: Learning to program in SPSS macros

It is true that Python and SPSS syntax are different languages, but it
is very powerful to use Python to generate SPSS commands as well as
using the direct functionality of Python function calls to SPSS.

But I'd like to clarify the "It's being suggested comment below". There
is no intention of replacing SPSS syntax with Python syntax entirely if
that is taken to mean that SPSS syntax would no longer be available.
What is interesting is the possibility of providing a pure Python set of
objects that would allow a Python programmer to write object-oriented
code that would generate and submit SPSS syntax underneath.

For example, you can open a data file and assign a dataset name to it
with SPSS syntax like this.

GET FILE='filespec'.
DATASET NAME mydataset.

Nothing wrong with that. But you could also accomplish this in a BEGIN
PROGRAM block with

spssaux.OpenDataFile('filespec', dataset='mydataset')

In similar ways you can wrap more complex SPSS syntax in Python
notation. The Transform module on Developer Central
(www.spss.com/devcentral) allows you to do COMPUTE and DO IF commands
along with VARIABLE LABEL, VARIABLE LEVEL, etc in o-o notation with
something like this.

This Python syntax defines all the metadata for a new variable, computes
it conditionally, and as an extra, makes the data updatable in the
future.

newvar = Transform.Compute(varname="average_increase",
varlabel="Salary increase per month of experience if at least a
year",\
varmeaslvl="Scale",\
varmissval=[999,998,997],\
varformat="F8.4")
newvar.expression = "(salary-salbegin)/jobtime"
newvar.condition = "jobtime > 12"
newvar.retransformable=True
newvar.generate() # An exception will be raised if the
compute fails

The equivalent SPSS syntax (without the updatable variable) would be
this.

NUMERIC average_increase.
FORMAT average_increase (F8.4).
VARIABLE LABEL average_increase "Salary increase per month of experience
if at least a year".
VARIABLE LEVEL average_increase (Scale).
MISSING VALUES average_increase (999,998, 997).
DO IF jobtime > 12.
COMPUTE average_increase = (salary-salbegin)/jobtime.
END IF.

So this approach turns SPSS syntax into a full fledged object-oriented
programming language, but it leaves SPSS syntax underneath doing the
work.

And it certainly does not eliminate the powerful SPSS syntax, which will
remain fully available. Not only is it widely known, there are zillions
of existing jobs that will continue to run. And new SPSS syntax will
continue to be created for new procedures. But if your style
inclinations are towards object-oriented programming, the Python
approach may appeal strongly. Someone, of course, would have to create
the Python code to generate all that SPSS syntax, which is not happening
now, at least within SPSS.

So this is an idea that I like, but that's my personal opinion.

Regards,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Richard Ristow
Sent: Wednesday, September 06, 2006 12:01 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Learning to program in SPSS macros

At 05:44 PM 9/5/2006, Sibusiso Moyo wrote:

>As a follow up question. How compatible is Python to SPSS? Can all the
>stuff that I do on SPSS be transferred to Python? Which is one more
>user friendly (in my company people like things simple :))? Can the
>Python code be easity re-integrated to SPSS syntax?

The short answer is, they're entirely different languages; in that
sense, not compatible at all. Nothing written to run in SPSS can run in
Python, or vice versa.

Python can create and submit SPSS statements that are then executed.
For how that's done; and for how different the syntaxes look; see my
just-now posting "Re: Renaming variables - noncontinuous", or various
postings by Jon Peck of SPSS, Inc.

It's being suggested that SPSS syntax be replaced with Python entirely;
that is, SPSS code will only be generated by Python programs, which will
use their own functions so that the users see none of the SPSS code. For
advocacy of that, see Jon Peck; for myself, I'm doubtful, for anything
like the near future. Among other things, it means a very careful set of
Python functions; and it means discarding much of the skill of current
SPSS programmers.

For more details, from a 'pro' point of view, see
http://www.spss.com/devcentral/files/articles/convert_spss_to_python1.pd
f

Peck, Jon

Re: Learning to program in SPSS macros

There is no official SPSS project at this point to "complete the mapping", but quite a lot has been done already, and the imminent SPSS 15 has further enhancements.

The Transform module that I described handles DO IF, COMPUTE, RECODE, and COUNT. These are the heavy duty transformation commands that do not have equivalents on the Python side. (None of this addresses INPUT PROGRAM). Something like, say, DO REPEAT, corresponds to iterating over a variable list, and this can already be written easily in Python code. In fact, the spssaux VariableDict class makes it easy to construct lists of variables by using any of lists of names, ranges (like TO but more general), filters on measurement level, type, etc, and wildcards on the names (every variable whose name includes AGE, for example).

If you construct variable lists this way or in more basic ways, then something like a hyper DO REPEAT is easy, and the Python loops could include procedures, too.

Now in SPSS 15, which is just being finalized now, there is another dramatic expansion in the ability to do SPSS transformations via Python. You can create new variables and assign properties, including the variable attributes introduced in SPSS 14 and values directly, so you can, in effect, do transformations in Python using any functions available or that you write that produce suitable casewise values.

I have just finished implementing a Soundex function, for example, and several others.

With SPSS 15, you can also append new cases.

So I believe that it is quite feasible with SPSS 15 to write almost any transformations in pure Python code running in or on top of SPSS if you so desire. I will be talking about this in more detail at the SPSS Directions conference in Chicago in early November (see www.spss.com for details on the conference).

Again let me emphasize that there is no plan to eliminate SPSS syntax or even an SPSS project to wrap everything in Python garb. And Python and the Plug-In are optional installs, so some users may not have access to these methods.

HTH,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bauer, Craig
Sent: Wednesday, September 06, 2006 9:55 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Learning to program in SPSS macros

Jon said:
"Someone, of course, would have to create the Python code to generate
all that SPSS syntax, which is not happening now, at least within SPSS."

Can you please tell us more about this? What % of transformation and
basic proc commands are currently mapped to python methods/properties
(e.g. could someone abandon SPSS code entirely now for simple things)?
Will SPSS be completing this mapping for future releases?

Thanks

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Peck, Jon
Sent: Wednesday, September 06, 2006 8:02 AM
To: [hidden email]
Subject: Re: Learning to program in SPSS macros

It is true that Python and SPSS syntax are different languages, but it
is very powerful to use Python to generate SPSS commands as well as
using the direct functionality of Python function calls to SPSS.

But I'd like to clarify the "It's being suggested comment below". There
is no intention of replacing SPSS syntax with Python syntax entirely if
that is taken to mean that SPSS syntax would no longer be available.
What is interesting is the possibility of providing a pure Python set of
objects that would allow a Python programmer to write object-oriented
code that would generate and submit SPSS syntax underneath.

For example, you can open a data file and assign a dataset name to it
with SPSS syntax like this.

GET FILE='filespec'.
DATASET NAME mydataset.

Nothing wrong with that. But you could also accomplish this in a BEGIN
PROGRAM block with

spssaux.OpenDataFile('filespec', dataset='mydataset')

In similar ways you can wrap more complex SPSS syntax in Python
notation. The Transform module on Developer Central
(www.spss.com/devcentral) allows you to do COMPUTE and DO IF commands
along with VARIABLE LABEL, VARIABLE LEVEL, etc in o-o notation with
something like this.

This Python syntax defines all the metadata for a new variable, computes
it conditionally, and as an extra, makes the data updatable in the
future.

newvar = Transform.Compute(varname="average_increase",
varlabel="Salary increase per month of experience if at least a
year",\
varmeaslvl="Scale",\
varmissval=[999,998,997],\
varformat="F8.4")
newvar.expression = "(salary-salbegin)/jobtime"
newvar.condition = "jobtime > 12"
newvar.retransformable=True
newvar.generate() # An exception will be raised if the
compute fails

The equivalent SPSS syntax (without the updatable variable) would be
this.

NUMERIC average_increase.
FORMAT average_increase (F8.4).
VARIABLE LABEL average_increase "Salary increase per month of experience
if at least a year".
VARIABLE LEVEL average_increase (Scale).
MISSING VALUES average_increase (999,998, 997).
DO IF jobtime > 12.
COMPUTE average_increase = (salary-salbegin)/jobtime.
END IF.

So this approach turns SPSS syntax into a full fledged object-oriented
programming language, but it leaves SPSS syntax underneath doing the
work.

And it certainly does not eliminate the powerful SPSS syntax, which will
remain fully available. Not only is it widely known, there are zillions
of existing jobs that will continue to run. And new SPSS syntax will
continue to be created for new procedures. But if your style
inclinations are towards object-oriented programming, the Python
approach may appeal strongly. Someone, of course, would have to create
the Python code to generate all that SPSS syntax, which is not happening
now, at least within SPSS.

So this is an idea that I like, but that's my personal opinion.

Regards,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Richard Ristow
Sent: Wednesday, September 06, 2006 12:01 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Learning to program in SPSS macros

At 05:44 PM 9/5/2006, Sibusiso Moyo wrote:

>As a follow up question. How compatible is Python to SPSS? Can all the
>stuff that I do on SPSS be transferred to Python? Which is one more
>user friendly (in my company people like things simple :))? Can the
>Python code be easity re-integrated to SPSS syntax?

The short answer is, they're entirely different languages; in that
sense, not compatible at all. Nothing written to run in SPSS can run in
Python, or vice versa.

Python can create and submit SPSS statements that are then executed.
For how that's done; and for how different the syntaxes look; see my
just-now posting "Re: Renaming variables - noncontinuous", or various
postings by Jon Peck of SPSS, Inc.

It's being suggested that SPSS syntax be replaced with Python entirely;
that is, SPSS code will only be generated by Python programs, which will
use their own functions so that the users see none of the SPSS code. For
advocacy of that, see Jon Peck; for myself, I'm doubtful, for anything
like the near future. Among other things, it means a very careful set of
Python functions; and it means discarding much of the skill of current
SPSS programmers.

For more details, from a 'pro' point of view, see
http://www.spss.com/devcentral/files/articles/convert_spss_to_python1.pd
f

Maguin, Eugene

Percentile calculations

All,

I've run an ancova and found that I have IV by covariate interaction. I then
did emmeans to locate the score at which the interaction became significant.
So I can say that at at score of 4.35 or above group A differed from group
B. What I'd like to do is turn that 4.35 into a percentile to give readers
an idea of how far out in the distribution that 4.35 actually is. So I look
at my distribution and I see that the cumulative percentage for a score of
4.3, which 103 people had, is 85.6% and the cum percentage for the next
higher score, 4.5, is 91.6%.

My question is how to state the percentile equivalent of 4.35 given the
observed distribution. Is the accepted practice to 'spread' the 103 people
over the interval between 85.6% and 91.6% and calculate the resulting
percentile as 87.1%? Or, is the accepted practice to do something else?

Thanks, Gene Maguin