SPSSX Discussion

PRINT lines outside MATRIX show inside MATRIX in output?

Classic

List

Threaded

12 messages Options

Marta García-Granero

PRINT lines outside MATRIX show inside MATRIX in output?

Hi everybody

Can you find an explanation for this behaviour?

If've written a macro for bootstraping Spearman's Rs correlation
coefficient. The arguments include the possibility of modifying the
seed and number of bootstrap samples (changing MXLOOPS to fit that
figure). The output uses PRINT to inform about the SEED status before
running MATRIX code for the bootstrap:

DEFINE RSICBOOT(vars=!TOKENS(2)/
k=!DEFAULT (10000) !TOKENS(1)/
seed=!DEFAULT(2541323) !TOKENS(1)).
PRESERVE.
SET MXLOOPS=!k.
SET SEED=!seed.
* Seed status report *.
DO IF $casenum EQ 1.
. PRINT.
. !IF (!UPCASE(!seed) !EQ 'RANDOM') !THEN.
. PRINT /'Semilla aleatoria (RANDOM)'.
. !ELSE.
. PRINT /'Valor de semilla: ' !QUOTE(!seed).
. !IFEND.
END IF.

* Matrix code starts here *.
MATRIX.
PRINT /TITLE="ESTIMACIÓN POR INTERVALO 'BOOTSTRAP' PARA Rs DE SPEARMAN".
COMPUTE k=!k.
GET data /VAR=!vars/MISSING=OMIT /NAME=vnames.
.
.

more code follows
.
.
END MATRIX.
RESTORE.
!ENDDEFINE.

As you can see, the PRINT statements are inside a DO IF and !IF
structures BEFORE the MATRIX code. Nonetheless, the printed lines
appear inside the MATRIX output, right after the title!:

This is the output:

*******************************
Run MATRIX procedure:

ESTIMACIÓN POR INTERVALO 'BOOTSTRAP' PARA Rs DE SPEARMAN

Valor de semilla: 2541323 <--- Here they are!!!

Variables analizadas
estriol peso

******* ESTADÍSTICOS DE LA MUESTRA ********

Coeficiente de correlación Rs de Spearman
Rs = .56

Tamaño muestral (n>10 para test de hipótesis fiable)
n = 31

Test de hipótesis para Rs (gl=n-2)
T 2*Sig
3.675 .001

******* RESULTADOS DE BOOTSTRAP ********

Condiciones de boostrap: k (Nº reps.)
10000

IC 95% para Rs (percentiles 2.5 y 97.5):
Inferior Superior
.25 .79

------ END MATRIX -----

*******************

It is not important (I don't mind having that infor displayed inside
the MATRIX output), but I'm just curious (I'm a scientist, it's in my
nature).

I hope this message reaches the list, I've been fighting with my
e-mail to try to send another message to the list for the whole day (I
don't who's to blame: the list or my mail services provider....).

Regards,
Marta

Zdaniuk, Bozena

transforming skewed var

Hello, I just read in an article that the authors conducted a
transformation of a skewed variable that was "...the opposite of the
inverse of the variable score". Could someone explain what kind of
transformation it is and whether it can be easily done in SPSS?
Thanks a lot.
Bozena

Bozena Zdaniuk, Ph.D.

University of Pittsburgh

UCSUR, 6th Fl.

121 University Place

Pittsburgh, PA 15260

Ph.: 412-624-5736

Fax: 412-624-4810

email: [hidden email]

Richard Ristow

Re: PRINT lines outside MATRIX show inside MATRIX in output?

In reply to this post by Marta García-Granero

At 10:33 AM 12/13/2006, Marta García-Granero wrote:

>Can you find an explanation for this behaviour?
>
>If've written a macro for bootstraping
>Spearman's Rs correlation coefficient. The
>output uses PRINT to inform about the SEED
>status before running MATRIX code for the bootstrap:
>
>DEFINE RSICBOOT(vars=!TOKENS(2)/
> k=!DEFAULT (10000) !TOKENS(1)/
> seed=!DEFAULT(2541323) !TOKENS(1)).
>PRESERVE.
>SET MXLOOPS=!k.
>SET SEED=!seed.
>* Seed status report *.
>DO IF $casenum EQ 1.
>. PRINT.
>. !IF (!UPCASE(!seed) !EQ 'RANDOM') !THEN.
>. PRINT /'Semilla aleatoria (RANDOM)'.
>. !ELSE.
>. PRINT /'Valor de semilla: ' !QUOTE(!seed).
>. !IFEND.
>END IF.
>
>* Matrix code starts here *.
>MATRIX.
>PRINT /TITLE="ESTIMACIÓN POR INTERVALO
>'BOOTSTRAP' PARA Rs DE SPEARMAN".
>COMPUTE k=!k.
>GET data /VAR=!vars/MISSING=OMIT /NAME=vnames.
>.
>more code follows
>.
>END MATRIX.
>RESTORE.
>!ENDDEFINE.
>
>As you can see, the PRINT statements are inside
>a DO IF and !IF structures BEFORE the MATRIX
>code. Nonetheless, the printed lines appear
>inside the MATRIX output, right after the title!:
>
>*******************************
>Run MATRIX procedure:
>
>ESTIMACIÓN POR INTERVALO 'BOOTSTRAP' PARA Rs DE SPEARMAN
>
>Valor de semilla: 2541323 <--- Here they are!!!

You are, I think, running into SPSS's
implementation that runs transformation programs
in parallel with the procedure or SAVE that
accepts their output. Doing it that way was one
of SPSS's best ideas, but it can be confusing
when the same transformation program writes printed output.

Your transformation program has only this code
(I'm not completely clear how far "SET SEED" acts
like a transformation-program executable):

SET SEED=!seed.
* Seed status report *.
DO IF $casenum EQ 1.
. PRINT.
. !IF (!UPCASE(!seed) !EQ 'RANDOM') !THEN.
. PRINT /'Semilla aleatoria (RANDOM)'.
. !ELSE.
. PRINT /'Valor de semilla: ' !QUOTE(!seed).
. !IFEND.
END IF.

Like any transformation program, its code is
read, interpreted, but NOT run until the
program's output is needed. The output is read in
the GET statement in your MATRIX code:

* Matrix code starts here *.
MATRIX.
PRINT /TITLE="ESTIMACIÓN POR INTERVALO
'BOOTSTRAP' PARA Rs DE SPEARMAN".
COMPUTE k=!k.
GET data /VAR=!vars/MISSING=OMIT /NAME=vnames.
...

So the sequence (in the code expanded from the macro) is,
1. The transformation program is read and
interpreted, and made ready to go

2. The MATRIX procedure begins, and announces that it's done so:
>Run MATRIX procedure:

3. The MATRIX program executes its initial PRINT statement, and prints
>ESTIMACIÓN POR INTERVALO 'BOOTSTRAP' PARA Rs DE SPEARMAN

4. The MATRIX program reads from the working file:
>GET data /VAR=!vars/MISSING=OMIT /NAME=vnames.

Then, and only then, SPSS begins to create the
working file that the GET is reading; and then,
and only then, executes the transformation
program, with the transformation-language PRINT statement.

You'd get the order you want if you force the
transformation program to run first, with an
EXECUTE before the MATRIX statement. I've
occasionally used EXECUTEs like that. In my
recent "Re: Converting international long
time-strings", Wed, 6 Dec 2006 (16:14:43 -0500),
the tracing code in the long transformation program includes
. /**/ EXECUTE.

so the output from the PRINT statements won't be
interspersed with the lines printed by the subsequent LIST procedure.

Cheers and great good wishes,
Richard

Hector Maletta

Re: transforming skewed var

In reply to this post by Zdaniuk, Bozena

Bozena,

If by "opposite" the authors meant the opposite sign, and by
"inverse" they meant the reciprocal, I do not see much rationality in it,
but SPSS could do it very easily; for any variable X the transformed
variable Y would be computed by:

COMPUTE Y=-(1/X).

If they did not mean this, I cannot immediately fathom what else
they could be talking about.

Besides, this mailing list has witnessed several debates on the
issue of transforming skewed or otherwise non-normally distributed variables
to obtain a transformed variable that is closer to normal. The general
consensus is that in most cases such transformations are neither advisable
nor necessary.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de
Zdaniuk, Bozena
Enviado el: 13 December 2006 16:41
Para: [hidden email]
Asunto: transforming skewed var

Hello, I just read in an article that the authors conducted a
transformation of a skewed variable that was "...the opposite of the
inverse of the variable score". Could someone explain what kind of
transformation it is and whether it can be easily done in SPSS?
Thanks a lot.
Bozena

Bozena Zdaniuk, Ph.D.

University of Pittsburgh

UCSUR, 6th Fl.

121 University Place

Pittsburgh, PA 15260

Ph.: 412-624-5736

Fax: 412-624-4810

email: [hidden email]

Zdaniuk, Bozena

Re: transforming skewed var

Thanks a lot, Hector. Very helpful, as always!
Bozena

Bozena Zdaniuk, Ph.D.

University of Pittsburgh

UCSUR, 6th Fl.

121 University Place

Pittsburgh, PA 15260

Ph.: 412-624-5736

Fax: 412-624-4810

email: [hidden email]

-----Original Message-----
From: Hector Maletta [mailto:[hidden email]]
Sent: Wednesday, December 13, 2006 1:57 PM
To: Zdaniuk, Bozena; [hidden email]
Subject: RE: transforming skewed var

Bozena,

If by "opposite" the authors meant the opposite sign, and by
"inverse" they meant the reciprocal, I do not see much rationality in
it,
but SPSS could do it very easily; for any variable X the transformed
variable Y would be computed by:

COMPUTE Y=-(1/X).

If they did not mean this, I cannot immediately fathom what else
they could be talking about.

Besides, this mailing list has witnessed several debates on the
issue of transforming skewed or otherwise non-normally distributed
variables
to obtain a transformed variable that is closer to normal. The general
consensus is that in most cases such transformations are neither
advisable
nor necessary.

Hector

-----Mensaje original-----
De: SPSSX(r) Discussion [mailto:[hidden email]] En nombre de
Zdaniuk, Bozena
Enviado el: 13 December 2006 16:41
Para: [hidden email]
Asunto: transforming skewed var

Hello, I just read in an article that the authors conducted a
transformation of a skewed variable that was "...the opposite of
the
inverse of the variable score". Could someone explain what kind
of
transformation it is and whether it can be easily done in SPSS?
Thanks a lot.
Bozena

Bozena Zdaniuk, Ph.D.

University of Pittsburgh

UCSUR, 6th Fl.

121 University Place

Pittsburgh, PA 15260

Ph.: 412-624-5736

Fax: 412-624-4810

email: [hidden email]

Richard Ristow

Re: PRINT lines outside MATRIX show inside MATRIX in output?

In reply to this post by Marta García-Granero

Postscript: at 10:33 AM 12/13/2006, Marta García-Granero wrote:

>I've written a macro for bootstrapping
>Spearman's Rs correlation coefficient. The
>output uses PRINT to inform about the SEED
>status before running MATRIX code for the bootstrap:

[code omitted]

>As you can see, the PRINT statements are inside
>a DO IF and !IF structures BEFORE the MATRIX
>code. Nonetheless, the printed lines appear
>inside the MATRIX output, right after the title!:

[output omitted]

When a macro simply 'wishes' to report about
itself and its arguments, ECHO is a good
alternative to a transformation program with
PRINT. ECHO is a pretty good substitute for the
!PRINT statement that might have been in the
macro facility. Simple implementation and test, below:

DEFINE RSICBOOT(vars=!TOKENS(2)/
k=!DEFAULT (10000) !TOKENS(1)/
seed=!DEFAULT(2541323) !TOKENS(1)).
PRESERVE.
SET MXLOOPS=!k.
SET SEED=!seed.
* Seed status report *.
!IF (!UPCASE(!seed) !EQ 'RANDOM') !THEN.
. ECHO 'Semilla aleatoria (RANDOM)'.
. !ELSE.
. ECHO !QUOTE(!CONCAT('Valor de semilla: ',!seed)).
. !IFEND.

* Matrix code starts here *.
MATRIX.
PRINT /TITLE="ESTIMACIÓN POR INTERVALO
'BOOTSTRAP' PARA Rs DE SPEARMAN".
COMPUTE k=!k.
GET data /VAR=!vars/MISSING=OMIT /NAME=vnames.
END MATRIX.
RESTORE.
!ENDDEFINE.

PRESERVE.
*..SET MPRINT ON.
RSICBOOT vars= a b.
Valor de semilla: 2541323

Run MATRIX procedure:

ESTIMACIÓN POR INTERVALO 'BOOTSTRAP' PARA Rs DE SPEARMAN

------ END MATRIX -----

RESTORE.

Albert-Jan Roskam

Re: transforming skewed var

In reply to this post by Zdaniuk, Bozena

Hi,

There's a handy one-page overview of commonly used
transformations in Stevens' "Applied multivariate
statistics for the social sciences'
(http://www.amazon.com/Applied-Multivariate-Statistics-Social-Sciences/dp/0805816704)

But I agree with Hector. Transformations rarely are
advisable. I remember I used a LN transformation once
for skewed data. Perhaps that's also an option for
you?

Cheers!
Albert-Jan

--- "Zdaniuk, Bozena" <[hidden email]> wrote:

> Thanks a lot, Hector. Very helpful, as always!
> Bozena
>
> Bozena Zdaniuk, Ph.D.
>
> University of Pittsburgh
>
> UCSUR, 6th Fl.
>
> 121 University Place
>
> Pittsburgh, PA 15260
>
> Ph.: 412-624-5736
>
> Fax: 412-624-4810
>
> email: [hidden email]
>
>
> -----Original Message-----
> From: Hector Maletta
> [mailto:[hidden email]]
> Sent: Wednesday, December 13, 2006 1:57 PM
> To: Zdaniuk, Bozena; [hidden email]
> Subject: RE: transforming skewed var
>
> Bozena,
>
> If by "opposite" the authors meant the
> opposite sign, and by
> "inverse" they meant the reciprocal, I do not see
> much rationality in
> it,
> but SPSS could do it very easily; for any variable X
> the transformed
> variable Y would be computed by:
>
> COMPUTE Y=-(1/X).
>
> If they did not mean this, I cannot
> immediately fathom what else
> they could be talking about.
>
> Besides, this mailing list has witnessed
> several debates on the
> issue of transforming skewed or otherwise
> non-normally distributed
> variables
> to obtain a transformed variable that is closer to
> normal. The general
> consensus is that in most cases such transformations
> are neither
> advisable
> nor necessary.
>
> Hector
>
>
>
> -----Mensaje original-----
> De: SPSSX(r) Discussion
> [mailto:[hidden email]] En nombre de
> Zdaniuk, Bozena
> Enviado el: 13 December 2006 16:41
> Para: [hidden email]
> Asunto: transforming skewed var
>
> Hello, I just read in an article that the
> authors conducted a
> transformation of a skewed variable that was
> "...the opposite of
> the
> inverse of the variable score". Could
> someone explain what kind
> of
> transformation it is and whether it can be
> easily done in SPSS?
> Thanks a lot.
> Bozena
>
> Bozena Zdaniuk, Ph.D.
>
> University of Pittsburgh
>
> UCSUR, 6th Fl.
>
> 121 University Place
>
> Pittsburgh, PA 15260
>
> Ph.: 412-624-5736
>
> Fax: 412-624-4810
>
> email: [hidden email]
>

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

statisticsdoc

Re: transforming skewed var

Stephen Brand
www.statisticsdoc.com

Log transformation of the dependent variable maybe particularly useful when
the dependent variable has a real (not arbitrary) zero-point (e.g., income).
In such cases, the dependent variable is measured as a ratio variable, and
it is meaningful to think of unit change in multiply/divide terms. Log
transformation may also be useful when the dependent variable is measured as
a ratio of one variable over another. In both cases, the raw distribution
of cases may be skewed. The transformation also serves a
variance-stabilizing function, such that the distribution of residuals is
similar across levels of predicted Y.

In the original example given by Dr. Zdaniuk, the investigator computed the
opposite of an inverse (as noted by Hector). The inverse of a dependent
variable (1/Y) is sometimes used as a variance stabilizing transformation
when the dependent variable represents time to some event (e.g., time to
die). (Taking the opposite does not change the shape of the distribution,
it merely reverses the direction of dependent variable). The square root
transformation of Y is also used as a variance stabilizing transformation
when Y is the result of Poisson counts.

A more complex transformation of Y, the arcsin of the square root of Y, may
be used when the Y is a proportion or a rate between 0 and 1.

Variable transformations have their uses, but consider whether the
transformed variables have meaning for your area of investigation (e.g., a
log transformation of a ratio-level variable like income might, a log
transformation of a likert scale probably would not). Also, consider
whether it might be more theoretically meaningful to add quadratic terms or
interactions between predictors to obtain better fit with the original raw
values of the dependent variable.

HTH,

Stephen Brand

For personalized and professional consultation in statistics and research
design, visit
www.statisticsdoc.com

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]]On Behalf Of
Albert-jan Roskam
Sent: Monday, December 18, 2006 4:11 AM
To: [hidden email]
Subject: Re: transforming skewed var

Hi,

There's a handy one-page overview of commonly used
transformations in Stevens' "Applied multivariate
statistics for the social sciences'
(http://www.amazon.com/Applied-Multivariate-Statistics-Social-Sciences/dp/08
05816704)

But I agree with Hector. Transformations rarely are
advisable. I remember I used a LN transformation once
for skewed data. Perhaps that's also an option for
you?

Cheers!
Albert-Jan

--- "Zdaniuk, Bozena" <[hidden email]> wrote:

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Gary Oliver

Missing values reported are incorrect

Colleagues

This is another SPSS v14.02 mystery for which I seek enlightenment. I
have a variable which has been set up to denote values from 1 to 7 both
inclusive. It has N=393.

Using SPSS menu and syntax for missing values check SPSS reports n=391
with 2 missing values. The peculiarity is that there are values in all
393 cells. All the values are 2-7 except for two cells which have values
of 1. The format is set to scale 4 places. When I ask SPSS to report the
missing cases it nominates the two cells which contain the value 1. I
looked in the help and there is no discussion of this.

Any suggestions?

Warm regards/gary

Unit Co-ordinator for Business Information Systems
(A Postgraduate Master of Commerce and Master of Business unit)
School of Business
The University of Sydney
------------------------
,-_|\ Building H69, Office 437
/ \ Corner Codrington Street
\_,-._* & Rose Street
Darlington 2006
@ Australia
--------------------------------------
E-mail: [hidden email]
------------------------
Location details:
Travelling from Broadway, turn south off City Road
Navigate toward the Acquatic Centre
------------------------
University Map: http://db.auth.usyd.edu.au/directories/map/index.stm
University Website:
www.usyd.edu.au
Faculty Website
www.econ.usyd.edu.au
------------------------
Faculty Student Information Office
(Timetables, Special Consideration)
Merewether Building
Enter from City Road side
e-mail: [hidden email]
Phone: 9351-3076
----------------------------------
Executive Officer for Business Information Systems
Katy Roy
Room 347, Building H69
E-mail: [hidden email]
Phone: 9036 9432
---------------------------------

>>-----Original Message-----
>>From: SPSSX(r) Discussion [mailto:[hidden email]]
>>On Behalf Of Statisticsdoc
>>Sent: Tuesday, December 19, 2006 12:18 PM
>>To: [hidden email]
>>Subject: Re: transforming skewed var
>>
>>Stephen Brand
>>www.statisticsdoc.com
>>
>>Log transformation of the dependent variable maybe
>>particularly useful when the dependent variable has a real
>>(not arbitrary) zero-point (e.g., income).
>>In such cases, the dependent variable is measured as a ratio
>>variable, and it is meaningful to think of unit change in
>>multiply/divide terms. Log transformation may also be useful
>>when the dependent variable is measured as a ratio of one
>>variable over another. In both cases, the raw distribution
>>of cases may be skewed. The transformation also serves a
>>variance-stabilizing function, such that the distribution of
>>residuals is similar across levels of predicted Y.
>>
>>In the original example given by Dr. Zdaniuk, the
>>investigator computed the opposite of an inverse (as noted by
>>Hector). The inverse of a dependent variable (1/Y) is
>>sometimes used as a variance stabilizing transformation when
>>the dependent variable represents time to some event (e.g.,
>>time to die). (Taking the opposite does not change the shape
>>of the distribution, it merely reverses the direction of
>>dependent variable). The square root transformation of Y is
>>also used as a variance stabilizing transformation when Y is
>>the result of Poisson counts.
>>
>>A more complex transformation of Y, the arcsin of the square
>>root of Y, may be used when the Y is a proportion or a rate
>>between 0 and 1.
>>
>>Variable transformations have their uses, but consider
>>whether the transformed variables have meaning for your area
>>of investigation (e.g., a log transformation of a ratio-level
>>variable like income might, a log transformation of a likert
>>scale probably would not). Also, consider whether it might
>>be more theoretically meaningful to add quadratic terms or
>>interactions between predictors to obtain better fit with
>>the original raw values of the dependent variable.
>>
>>HTH,
>>
>>Stephen Brand
>>
>>
>>For personalized and professional consultation in statistics
>>and research design, visit www.statisticsdoc.com
>>
>>
>>-----Original Message-----
>>From: SPSSX(r) Discussion [mailto:[hidden email]]On
>>Behalf Of Albert-jan Roskam
>>Sent: Monday, December 18, 2006 4:11 AM
>>To: [hidden email]
>>Subject: Re: transforming skewed var
>>
>>
>>Hi,
>>
>>There's a handy one-page overview of commonly used
>>transformations in Stevens' "Applied multivariate statistics
>>for the social sciences'
>>(http://www.amazon.com/Applied-Multivariate-Statistics-Social-
>>Sciences/dp/08
>>05816704)
>>
>>But I agree with Hector. Transformations rarely are
>>advisable. I remember I used a LN transformation once for
>>skewed data. Perhaps that's also an option for you?
>>
>>Cheers!
>>Albert-Jan
>>
>>--- "Zdaniuk, Bozena" <[hidden email]> wrote:
>>
>>> Thanks a lot, Hector. Very helpful, as always!
>>> Bozena
>>>
>>> Bozena Zdaniuk, Ph.D.
>>>
>>> University of Pittsburgh
>>>
>>> UCSUR, 6th Fl.
>>>
>>> 121 University Place
>>>
>>> Pittsburgh, PA 15260
>>>
>>> Ph.: 412-624-5736
>>>
>>> Fax: 412-624-4810
>>>
>>> email: [hidden email]
>>>
>>>
>>> -----Original Message-----
>>> From: Hector Maletta
>>> [mailto:[hidden email]]
>>> Sent: Wednesday, December 13, 2006 1:57 PM
>>> To: Zdaniuk, Bozena; [hidden email]
>>> Subject: RE: transforming skewed var
>>>
>>> Bozena,
>>>
>>> If by "opposite" the authors meant the opposite
>>sign, and by
>>> "inverse" they meant the reciprocal, I do not see much
>>rationality in
>>> it, but SPSS could do it very easily; for any variable X the
>>> transformed variable Y would be computed by:
>>>
>>> COMPUTE Y=-(1/X).
>>>
>>> If they did not mean this, I cannot immediately fathom what
>>> else they could be talking about.
>>>
>>> Besides, this mailing list has witnessed several debates on
>>> the issue of transforming skewed or otherwise non-normally
>>distributed
>>> variables to obtain a transformed variable that is closer
>>to normal.
>>> The general consensus is that in most cases such
>>transformations are
>>> neither advisable nor necessary.
>>>
>>> Hector
>>>
>>>
>>>
>>> -----Mensaje original-----
>>> De: SPSSX(r) Discussion
>>> [mailto:[hidden email]] En nombre de Zdaniuk,
>>Bozena Enviado
>>> el: 13 December 2006 16:41
>>> Para: [hidden email]
>>> Asunto: transforming skewed var
>>>
>>> Hello, I just read in an article that the authors
>>conducted a
>>> transformation of a skewed variable that was
>>"...the opposite
>>> of the
>>> inverse of the variable score". Could someone explain what
>>> kind of
>>> transformation it is and whether it can be easily done in
>>> SPSS?
>>> Thanks a lot.
>>> Bozena
>>>
>>> Bozena Zdaniuk, Ph.D.
>>>
>>> University of Pittsburgh
>>>
>>> UCSUR, 6th Fl.
>>>
>>> 121 University Place
>>>
>>> Pittsburgh, PA 15260
>>>
>>> Ph.: 412-624-5736
>>>
>>> Fax: 412-624-4810
>>>
>>> email: [hidden email]
>>>
>>
>>
>>__________________________________________________
>>Do You Yahoo!?
>>Tired of spam? Yahoo! Mail has the best spam protection
>>around http://mail.yahoo.com
>>

Spousta Jan

Re: Missing values reported are incorrect

Looks like 1 is declared as a user missing value. Run

miss val here_insert_the_variable_name ().

from the syntax editor (or edit the Missing column in the Data Editor,
Variable View) and then re-run frequencies.

HTH

Jan

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Gary Oliver
Sent: Wednesday, December 20, 2006 7:53 AM
To: [hidden email]
Subject: Missing values reported are incorrect

Colleagues

This is another SPSS v14.02 mystery for which I seek enlightenment. I
have a variable which has been set up to denote values from 1 to 7 both
inclusive. It has N=393.

Using SPSS menu and syntax for missing values check SPSS reports n=391
with 2 missing values. The peculiarity is that there are values in all
393 cells. All the values are 2-7 except for two cells which have values
of 1. The format is set to scale 4 places. When I ask SPSS to report the
missing cases it nominates the two cells which contain the value 1. I
looked in the help and there is no discussion of this.

Any suggestions?

Warm regards/gary

Unit Co-ordinator for Business Information Systems (A Postgraduate
Master of Commerce and Master of Business unit) School of Business The
University of Sydney
------------------------
,-_|\ Building H69, Office 437
/ \ Corner Codrington Street
\_,-._* & Rose Street
Darlington 2006
@ Australia
--------------------------------------
E-mail: [hidden email]
------------------------
Location details:
Travelling from Broadway, turn south off City Road Navigate toward the
Acquatic Centre
------------------------
University Map: http://db.auth.usyd.edu.au/directories/map/index.stm
University Website:
www.usyd.edu.au
Faculty Website
www.econ.usyd.edu.au
------------------------
Faculty Student Information Office
(Timetables, Special Consideration)
Merewether Building
Enter from City Road side
e-mail: [hidden email]
Phone: 9351-3076
----------------------------------
Executive Officer for Business Information Systems Katy Roy Room 347,
Building H69
E-mail: [hidden email]
Phone: 9036 9432
---------------------------------

>>-----Original Message-----
>>From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf
>>Of Statisticsdoc
>>Sent: Tuesday, December 19, 2006 12:18 PM
>>To: [hidden email]
>>Subject: Re: transforming skewed var
>>
>>Stephen Brand
>>www.statisticsdoc.com
>>
>>Log transformation of the dependent variable maybe particularly useful

>>when the dependent variable has a real (not arbitrary) zero-point
>>(e.g., income).
>>In such cases, the dependent variable is measured as a ratio variable,

>>and it is meaningful to think of unit change in multiply/divide terms.

>>Log transformation may also be useful when the dependent variable is
>>measured as a ratio of one variable over another. In both cases, the
>>raw distribution of cases may be skewed. The transformation also
>>serves a variance-stabilizing function, such that the distribution of
>>residuals is similar across levels of predicted Y.
>>
>>In the original example given by Dr. Zdaniuk, the investigator
>>computed the opposite of an inverse (as noted by Hector). The inverse

>>of a dependent variable (1/Y) is sometimes used as a variance
>>stabilizing transformation when the dependent variable represents time

>>to some event (e.g., time to die). (Taking the opposite does not
>>change the shape of the distribution, it merely reverses the direction

>>of dependent variable). The square root transformation of Y is also
>>used as a variance stabilizing transformation when Y is the result of
>>Poisson counts.
>>
>>A more complex transformation of Y, the arcsin of the square root of
>>Y, may be used when the Y is a proportion or a rate between 0 and 1.
>>
>>Variable transformations have their uses, but consider whether the
>>transformed variables have meaning for your area of investigation
>>(e.g., a log transformation of a ratio-level variable like income
>>might, a log transformation of a likert scale probably would not).
>>Also, consider whether it might be more theoretically meaningful to
>>add quadratic terms or interactions between predictors to obtain
>>better fit with the original raw values of the dependent variable.
>>
>>HTH,
>>
>>Stephen Brand
>>
>>
>>For personalized and professional consultation in statistics and
>>research design, visit www.statisticsdoc.com
>>
>>
>>-----Original Message-----
>>From: SPSSX(r) Discussion [mailto:[hidden email]]On
>>Behalf Of Albert-jan Roskam
>>Sent: Monday, December 18, 2006 4:11 AM
>>To: [hidden email]
>>Subject: Re: transforming skewed var
>>
>>
>>Hi,
>>
>>There's a handy one-page overview of commonly used transformations in
>>Stevens' "Applied multivariate statistics for the social sciences'
>>(http://www.amazon.com/Applied-Multivariate-Statistics-Social-
>>Sciences/dp/08
>>05816704)
>>
>>But I agree with Hector. Transformations rarely are advisable. I
>>remember I used a LN transformation once for skewed data. Perhaps
>>that's also an option for you?
>>
>>Cheers!
>>Albert-Jan
>>
>>--- "Zdaniuk, Bozena" <[hidden email]> wrote:
>>
>>> Thanks a lot, Hector. Very helpful, as always!
>>> Bozena
>>>
>>> Bozena Zdaniuk, Ph.D.
>>>
>>> University of Pittsburgh
>>>
>>> UCSUR, 6th Fl.
>>>
>>> 121 University Place
>>>
>>> Pittsburgh, PA 15260
>>>
>>> Ph.: 412-624-5736
>>>
>>> Fax: 412-624-4810
>>>
>>> email: [hidden email]
>>>
>>>
>>> -----Original Message-----
>>> From: Hector Maletta
>>> [mailto:[hidden email]]
>>> Sent: Wednesday, December 13, 2006 1:57 PM
>>> To: Zdaniuk, Bozena; [hidden email]
>>> Subject: RE: transforming skewed var
>>>
>>> Bozena,
>>>
>>> If by "opposite" the authors meant the opposite
>>sign, and by
>>> "inverse" they meant the reciprocal, I do not see much
>>rationality in
>>> it, but SPSS could do it very easily; for any variable X the
>>> transformed variable Y would be computed by:
>>>
>>> COMPUTE Y=-(1/X).
>>>
>>> If they did not mean this, I cannot immediately fathom what
>>> else they could be talking about.
>>>
>>> Besides, this mailing list has witnessed several debates on
>>> the issue of transforming skewed or otherwise non-normally
>>distributed
>>> variables to obtain a transformed variable that is closer
>>to normal.
>>> The general consensus is that in most cases such
>>transformations are
>>> neither advisable nor necessary.
>>>
>>> Hector
>>>
>>>
>>>
>>> -----Mensaje original-----
>>> De: SPSSX(r) Discussion
>>> [mailto:[hidden email]] En nombre de Zdaniuk,
>>Bozena Enviado
>>> el: 13 December 2006 16:41
>>> Para: [hidden email]
>>> Asunto: transforming skewed var
>>>
>>> Hello, I just read in an article that the authors
>>conducted a
>>> transformation of a skewed variable that was
>>"...the opposite
>>> of the
>>> inverse of the variable score". Could someone explain what
>>> kind of
>>> transformation it is and whether it can be easily done in
>>> SPSS?
>>> Thanks a lot.
>>> Bozena
>>>
>>> Bozena Zdaniuk, Ph.D.
>>>
>>> University of Pittsburgh
>>>
>>> UCSUR, 6th Fl.
>>>
>>> 121 University Place
>>>
>>> Pittsburgh, PA 15260
>>>
>>> Ph.: 412-624-5736
>>>
>>> Fax: 412-624-4810
>>>
>>> email: [hidden email]
>>>
>>
>>
>>__________________________________________________
>>Do You Yahoo!?
>>Tired of spam? Yahoo! Mail has the best spam protection around
>>http://mail.yahoo.com
>>

Melissa Ives

Re: Missing values reported are incorrect

In reply to this post by Gary Oliver

To see the currently assigned missing values, go to the variable view
and see what is entered in the "Missing" column for your variable. If
it is something other than "None", then you can open the dialog box for
missing and select 'No missing values' or you could copy "None" from
another variable's 'Missing' cell and paste it into this variable's
'Missing' cell.

Melissa

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Spousta Jan
Sent: Wednesday, December 20, 2006 7:33 AM
To: [hidden email]
Subject: Re: [SPSSX-L] Missing values reported are incorrect

Looks like 1 is declared as a user missing value. Run

miss val here_insert_the_variable_name ().

from the syntax editor (or edit the Missing column in the Data Editor,
Variable View) and then re-run frequencies.

HTH

Jan

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Gary Oliver
Sent: Wednesday, December 20, 2006 7:53 AM
To: [hidden email]
Subject: Missing values reported are incorrect

Colleagues

This is another SPSS v14.02 mystery for which I seek enlightenment. I
have a variable which has been set up to denote values from 1 to 7 both
inclusive. It has N=393.

Using SPSS menu and syntax for missing values check SPSS reports n=391
with 2 missing values. The peculiarity is that there are values in all
393 cells. All the values are 2-7 except for two cells which have values
of 1. The format is set to scale 4 places. When I ask SPSS to report the
missing cases it nominates the two cells which contain the value 1. I
looked in the help and there is no discussion of this.

Any suggestions?

Warm regards/gary

Unit Co-ordinator for Business Information Systems (A Postgraduate
Master of Commerce and Master of Business unit) School of Business The
University of Sydney
------------------------
,-_|\ Building H69, Office 437
/ \ Corner Codrington Street
\_,-._* & Rose Street
Darlington 2006
@ Australia
--------------------------------------
E-mail: [hidden email]
------------------------
Location details:
Travelling from Broadway, turn south off City Road Navigate toward the
Acquatic Centre
------------------------
University Map: http://db.auth.usyd.edu.au/directories/map/index.stm
University Website:
www.usyd.edu.au
Faculty Website
www.econ.usyd.edu.au
------------------------
Faculty Student Information Office
(Timetables, Special Consideration)
Merewether Building
Enter from City Road side
e-mail: [hidden email]
Phone: 9351-3076
----------------------------------
Executive Officer for Business Information Systems Katy Roy Room 347,
Building H69
E-mail: [hidden email]
Phone: 9036 9432
---------------------------------

PRIVILEGED AND CONFIDENTIAL INFORMATION
This transmittal and any attachments may contain PRIVILEGED AND
CONFIDENTIAL information and is intended only for the use of the
addressee. If you are not the designated recipient, or an employee
or agent authorized to deliver such transmittals to the designated
recipient, you are hereby notified that any dissemination,
copying or publication of this transmittal is strictly prohibited. If
you have received this transmittal in error, please notify us
immediately by replying to the sender and delete this copy from your
system. You may also call us at (309) 827-6026 for assistance.

Brian Pickerill

Re: Missing values reported are incorrect

In reply to this post by Gary Oliver

You didn't really say which values you wanted to be set as (user)
missing. Assuming you want 2 thru 7 missing, try this in syntax:

missing values YOUR_VARIABLE_NAME (2 thru 7).

I don't know about v14 yet (I use Macs) but you can generally list
up to 3 values within the parentheses, or use the 'thru' keyword
as above to give an inclusive range of values.)

No need to clear the missing values first.

--Brian

On Dec 20, 2006, at 1:52 AM, Gary Oliver wrote:

> Colleagues
>
> This is another SPSS v14.02 mystery for which I seek enlightenment. I
> have a variable which has been set up to denote values from 1 to 7
> both
> inclusive. It has N=393.
>
> Using SPSS menu and syntax for missing values check SPSS reports n=391
> with 2 missing values. The peculiarity is that there are values in all
> 393 cells. All the values are 2-7 except for two cells which have
> values
> of 1. The format is set to scale 4 places. When I ask SPSS to
> report the
> missing cases it nominates the two cells which contain the value 1. I
> looked in the help and there is no discussion of this.
>
> Any suggestions?
>
> Warm regards/gary
>
> Unit Co-ordinator for Business Information Systems
> (A Postgraduate Master of Commerce and Master of Business unit)
> School of Business
> The University of Sydney
> ------------------------
> ,-_|\ Building H69, Office 437
> / \ Corner Codrington Street
> \_,-._* & Rose Street
> Darlington 2006
> @ Australia
> --------------------------------------
> E-mail: [hidden email]
> ------------------------
> Location details:
> Travelling from Broadway, turn south off City Road
> Navigate toward the Acquatic Centre
> ------------------------
> University Map: http://db.auth.usyd.edu.au/directories/map/index.stm
> University Website:
> www.usyd.edu.au
> Faculty Website
> www.econ.usyd.edu.au
> ------------------------
> Faculty Student Information Office
> (Timetables, Special Consideration)
> Merewether Building
> Enter from City Road side
> e-mail: [hidden email]
> Phone: 9351-3076
> ----------------------------------
> Executive Officer for Business Information Systems
> Katy Roy
> Room 347, Building H69
> E-mail: [hidden email]
> Phone: 9036 9432
> ---------------------------------
>
>
>>> -----Original Message-----
>>> From: SPSSX(r) Discussion [mailto:[hidden email]]
>>> On Behalf Of Statisticsdoc
>>> Sent: Tuesday, December 19, 2006 12:18 PM
>>> To: [hidden email]
>>> Subject: Re: transforming skewed var
>>>
>>> Stephen Brand
>>> www.statisticsdoc.com
>>>
>>> Log transformation of the dependent variable maybe
>>> particularly useful when the dependent variable has a real
>>> (not arbitrary) zero-point (e.g., income).
>>> In such cases, the dependent variable is measured as a ratio
>>> variable, and it is meaningful to think of unit change in
>>> multiply/divide terms. Log transformation may also be useful
>>> when the dependent variable is measured as a ratio of one
>>> variable over another. In both cases, the raw distribution
>>> of cases may be skewed. The transformation also serves a
>>> variance-stabilizing function, such that the distribution of
>>> residuals is similar across levels of predicted Y.
>>>
>>> In the original example given by Dr. Zdaniuk, the
>>> investigator computed the opposite of an inverse (as noted by
>>> Hector). The inverse of a dependent variable (1/Y) is
>>> sometimes used as a variance stabilizing transformation when
>>> the dependent variable represents time to some event (e.g.,
>>> time to die). (Taking the opposite does not change the shape
>>> of the distribution, it merely reverses the direction of
>>> dependent variable). The square root transformation of Y is
>>> also used as a variance stabilizing transformation when Y is
>>> the result of Poisson counts.
>>>
>>> A more complex transformation of Y, the arcsin of the square
>>> root of Y, may be used when the Y is a proportion or a rate
>>> between 0 and 1.
>>>
>>> Variable transformations have their uses, but consider
>>> whether the transformed variables have meaning for your area
>>> of investigation (e.g., a log transformation of a ratio-level
>>> variable like income might, a log transformation of a likert
>>> scale probably would not). Also, consider whether it might
>>> be more theoretically meaningful to add quadratic terms or
>>> interactions between predictors to obtain better fit with
>>> the original raw values of the dependent variable.
>>>
>>> HTH,
>>>
>>> Stephen Brand
>>>
>>>
>>> For personalized and professional consultation in statistics
>>> and research design, visit www.statisticsdoc.com
>>>
>>>
>>> -----Original Message-----
>>> From: SPSSX(r) Discussion [mailto:[hidden email]]On
>>> Behalf Of Albert-jan Roskam
>>> Sent: Monday, December 18, 2006 4:11 AM
>>> To: [hidden email]
>>> Subject: Re: transforming skewed var
>>>
>>>
>>> Hi,
>>>
>>> There's a handy one-page overview of commonly used
>>> transformations in Stevens' "Applied multivariate statistics
>>> for the social sciences'
>>> (http://www.amazon.com/Applied-Multivariate-Statistics-Social-
>>> Sciences/dp/08
>>> 05816704)
>>>
>>> But I agree with Hector. Transformations rarely are
>>> advisable. I remember I used a LN transformation once for
>>> skewed data. Perhaps that's also an option for you?
>>>
>>> Cheers!
>>> Albert-Jan
>>>
>>> --- "Zdaniuk, Bozena" <[hidden email]> wrote:
>>>
>>>> Thanks a lot, Hector. Very helpful, as always!
>>>> Bozena
>>>>
>>>> Bozena Zdaniuk, Ph.D.
>>>>
>>>> University of Pittsburgh
>>>>
>>>> UCSUR, 6th Fl.
>>>>
>>>> 121 University Place
>>>>
>>>> Pittsburgh, PA 15260
>>>>
>>>> Ph.: 412-624-5736
>>>>
>>>> Fax: 412-624-4810
>>>>
>>>> email: [hidden email]
>>>>
>>>>
>>>> -----Original Message-----
>>>> From: Hector Maletta
>>>> [mailto:[hidden email]]
>>>> Sent: Wednesday, December 13, 2006 1:57 PM
>>>> To: Zdaniuk, Bozena; [hidden email]
>>>> Subject: RE: transforming skewed var
>>>>
>>>> Bozena,
>>>>
>>>> If by "opposite" the authors meant the opposite
>>> sign, and by
>>>> "inverse" they meant the reciprocal, I do not see much
>>> rationality in
>>>> it, but SPSS could do it very easily; for any variable X the
>>>> transformed variable Y would be computed by:
>>>>
>>>> COMPUTE Y=-(1/X).
>>>>
>>>> If they did not mean this, I cannot immediately fathom what
>>>> else they could be talking about.
>>>>
>>>> Besides, this mailing list has witnessed several debates on
>>>> the issue of transforming skewed or otherwise non-normally
>>> distributed
>>>> variables to obtain a transformed variable that is closer
>>> to normal.
>>>> The general consensus is that in most cases such
>>> transformations are
>>>> neither advisable nor necessary.
>>>>
>>>> Hector
>>>>
>>>>
>>>>
>>>> -----Mensaje original-----
>>>> De: SPSSX(r) Discussion
>>>> [mailto:[hidden email]] En nombre de Zdaniuk,
>>> Bozena Enviado
>>>> el: 13 December 2006 16:41
>>>> Para: [hidden email]
>>>> Asunto: transforming skewed var
>>>>
>>>> Hello, I just read in an article that the authors
>>> conducted a
>>>> transformation of a skewed variable that was
>>> "...the opposite
>>>> of the
>>>> inverse of the variable score". Could someone explain what
>>>> kind of
>>>> transformation it is and whether it can be easily done in
>>>> SPSS?
>>>> Thanks a lot.
>>>> Bozena
>>>>
>>>> Bozena Zdaniuk, Ph.D.
>>>>
>>>> University of Pittsburgh
>>>>
>>>> UCSUR, 6th Fl.
>>>>
>>>> 121 University Place
>>>>
>>>> Pittsburgh, PA 15260
>>>>
>>>> Ph.: 412-624-5736
>>>>
>>>> Fax: 412-624-4810
>>>>
>>>> email: [hidden email]
>>>>
>>>
>>>
>>> __________________________________________________
>>> Do You Yahoo!?
>>> Tired of spam? Yahoo! Mail has the best spam protection
>>> around http://mail.yahoo.com
>>>