use saved coefficients from a regression in the data file

[drfg2008]

SPSS 17 / Vista

I would like to use the coefficients, which were saved from a regression in a sav-file,  within a new data file.

Here the example:

http://www.frag-einen-statistiker.de/spss/koeffizientenmatrix.sav

for example:

compute pb21lay_c = 1,194 * standard_lay2 + ... + 1,112                                                                                  

plus all the other variables (8 variables alltogether)


couldn't find a solution.

Thanks

[Jon K Peck]

Since you have already captured the coefficients, you could use MATRIX for this.
Alternatively, read the dataset with programmability and generate the required COMPUTE statements from it and use spss.Submit to run them.

Jon Peck (no "h")
Senior Software Engineer, IBM
[hidden email]
new phone: [phone]




From:        drfg2008 <[hidden email]>
To:        [hidden email]
Date:        08/20/2011 01:58 PM
Subject:        [SPSSX-L] use saved coefficients from a regression in the data              file
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

SPSS 17 / Vista

I would like to use the coefficients, which were saved from a regression in
a sav-file,  within a new data file.

Here the example:

http://www.frag-einen-statistiker.de/spss/koeffizientenmatrix.sav

for example:

compute pb21lay_c = 1,194 * standard_lay2 + ... + 1,112

plus all the other variables (8 variables alltogether)


couldn't find a solution.

Thanks

-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/use-saved-coefficients-from-a-regression-in-the-data-file-tp4719101p4719101.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[drfg2008]

read the dataset with programmability


The only way I figured out so far is to use means tables [1] or CTABLES [2] (to read the data) and combine it with python [3]. Is that what is meant by programmability? Or is there a more straight forward way to it.


Frank

[1]
MEANS TABLES=pb02back_c pb02lay_c pb12back_c pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c BY  term
  /CELLS MEAN.

[2]
CTABLES
  /VLABELS VARIABLES=term DISPLAY=DEFAULT  /VLABELS VARIABLES=pb02back_c pb02lay_c pb12back_c
    pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c
    DISPLAY=BOTH
  /TABLE term BY pb02back_c [MEAN] + pb02lay_c [MEAN] + pb12back_c [MEAN] + pb12lay_c [MEAN] +
    pb20back_c [MEAN] + pb20lay_c [MEAN] + pb21back_c [MEAN] + pb21lay_c [MEAN]
  /CATEGORIES VARIABLES=term ORDER=A KEY=VALUE EMPTY=EXCLUDE.


[3]
desc_table,errcode=spssaux.CreateXMLOutput(cmd, omsid= ...

[David Marso]

Frank,
I have absolutely NO idea what you are up here to with CTABLES .... and OMS??  WTF is the point of that?
Jon suggested MATRIX.
I suggest you try his suggestion. ;-)
Say your coefficients are in some SPSS system file Beta.sav .(Px1)
Your Independent vars are in a SPSS system file RawX.sav . (NxP)
UNTESTED off the cuff code -and it probably WORKS-.

You may need to add a column of 1's to Raw.Sav for constant term.
You can do that in MATRIX or outside.  May need to alter the !VARS list accordingly.

DEFINE !VARS ()
pb02back_c pb02lay_c pb12back_c pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c
!ENDDEFINE .

MATRIX.
GET X /VARIABLES !VARS / FILE "RawX.sav".
GET Beta / VARIABLES !VARS / FILE "Beta.sav".
COMPUTE Pred=X*Beta.
COMPUTE SNames={ !VARS, Predicted }.
SAVE { X, Pred } / NAMES= SNames/ OUTFILE *.
END MATRIX.

Hey Look PEEPS, No freakin Python!!!!!
HTH, David





--

drfg2008 wrote

read the dataset with programmability


The only way I figured out so far is to use means tables [1] or CTABLES [2] (to read the data) and combine it with python [3]. Is that what is meant by programmability? Or is there a more straight forward way to it.


Frank

[1]
MEANS TABLES=pb02back_c pb02lay_c pb12back_c pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c BY  term
  /CELLS MEAN.

[2]
CTABLES
  /VLABELS VARIABLES=term DISPLAY=DEFAULT  /VLABELS VARIABLES=pb02back_c pb02lay_c pb12back_c
    pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c
    DISPLAY=BOTH
  /TABLE term BY pb02back_c [MEAN] + pb02lay_c [MEAN] + pb12back_c [MEAN] + pb12lay_c [MEAN] +
    pb20back_c [MEAN] + pb20lay_c [MEAN] + pb21back_c [MEAN] + pb21lay_c [MEAN]
  /CATEGORIES VARIABLES=term ORDER=A KEY=VALUE EMPTY=EXCLUDE.


[3]
desc_table,errcode=spssaux.CreateXMLOutput(cmd, omsid= ...

[Jon K Peck]

MEANS and CTABLES don't have anything to do with the solutions I suggested.  You can save the regression coefficients using either OMS or REGRESSION command syntax ( /OUTFILE=COVB(coefs).)

You can read that into MATRIX along with the main dataset, extract the relevant columns, and use matrix operations to create the product.

Or you could write Python code to read the coefficient matrix and generate transformation commands.  But you don't need to do that here.

HTH,

Jon Peck (no "h")
Senior Software Engineer, IBM
[hidden email]
new phone: [phone]




From:        drfg2008 <[hidden email]>
To:        [hidden email]
Date:        08/23/2011 06:14 AM
Subject:        Re: [SPSSX-L] use saved coefficients from a regression in the              data file
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

/read the dataset with programmability/


The only way I figured out so far is to use means tables [1] or CTABLES [2]
(to read the data) and combine it with python [3]. Is that what is meant by
programmability? Or is there a more straight forward way to it.


Frank

[1]
MEANS TABLES=pb02back_c pb02lay_c pb12back_c pb12lay_c pb20back_c pb20lay_c
pb21back_c pb21lay_c BY  term
 /CELLS MEAN.

[2]
CTABLES
 /VLABELS VARIABLES=term DISPLAY=DEFAULT  /VLABELS VARIABLES=pb02back_c
pb02lay_c pb12back_c
   pb12lay_c pb20back_c pb20lay_c pb21back_c pb21lay_c
   DISPLAY=BOTH
 /TABLE term BY pb02back_c [MEAN] + pb02lay_c [MEAN] + pb12back_c [MEAN] +
pb12lay_c [MEAN] +
   pb20back_c [MEAN] + pb20lay_c [MEAN] + pb21back_c [MEAN] + pb21lay_c
[MEAN]
 /CATEGORIES VARIABLES=term ORDER=A KEY=VALUE EMPTY=EXCLUDE.


[3]
desc_table,errcode=spssaux.CreateXMLOutput(cmd, omsid= ...



-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/use-saved-coefficients-from-a-regression-in-the-data-file-tp4719101p4726551.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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]

Another way  (in addition to matrix or computes) to get the predicted Y from a regression
add the two files with /IN = set1
and run the regression using /SELECT set1 EQ 1.

Art Kendall
Social Research Consultants

On 8/20/2011 3:54 PM, drfg2008 wrote:

SPSS 17 / Vista

I would like to use the coefficients, which were saved from a regression in
a sav-file,  within a new data file.

Here the example:

http://www.frag-einen-statistiker.de/spss/koeffizientenmatrix.sav

for example:

compute pb21lay_c = 1,194 * standard_lay2 + ... + 1,112

plus all the other variables (8 variables alltogether)


couldn't find a solution.

Thanks

-----
Dr. Frank Gaeth
FU-Berlin

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/use-saved-coefficients-from-a-regression-in-the-data-file-tp4719101p4719101.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
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

[David Marso]

OK!!! Time Out Frank,
I finally booted into Windows where I have SPSS installed and opened your file.
You could just have easily copy/pasted it in your original message.
term pb02back pb02lay pb12back pb12lay pb20back pb20lay pb21back pb21lay
(Intercept)              -1.028    -1.086     1.043     1.258    -1.010    -1.099      .808     1.112
standard_Back_1       .000      .000     1.011    -1.084     4.313      .000     1.677     2.044
standard_Lay_1        -.001     -.005     -.694     1.321    -2.167     2.257     -.009     -.149
standard_Back_2       5.051      .179      .006      .787      .005     -.003      .659     -.941
standard_Lay_2       -2.882     2.077     1.556     1.038     -.005      .000     -.329     1.194

My previous MATRIX solution stands except use your terms as elements of !VAR and add the dependent variable names (pb02back pb02lay ...pb21lay) on the SNAMES.
Don't forget to create the vector of ones called Intercept  in the RawX.sav file.

DEFINE !VARS ()
Intercept standard_Back_1 standard_Lay_1 standard_Back_2 standard_Lay_2
!ENDDEFINE .

MATRIX.
GET X /VARIABLES !VARS / FILE "RawX.sav".
GET Beta / VARIABLES !VARS / FILE "Beta.sav".
COMPUTE Pred=X*Beta.
COMPUTE SNames={ !VARS, "pb02back_c", "pb02lay_c", "pb12back_c", "pb12lay_c" ,
                                             "pb20back_c", "pb20lay_c", "pb21back_c", "pb21lay_c" }.
SAVE { X, Pred } / NAMES= SNames/ OUTFILE *.
END MATRIX.

Hey Look PEEPS, No freakin Python!!!!!
HTH, David