Creating a list of significant variables

[Bryan Tec]

Hi all,



I am doing some exploratory work involving adding variables one at a time to
a base regression model.  There is a large number of variables to be added
one at a time and hence a large number of models to be run.  Rather than
manually going through them to find the significant results I would like to
have SPSS create a list of the variables that were significant.  For
example:



DV: Y

Standard IVs: A and B

List of IVs to be added one at a time: C, D, E, F.



Model 1:



Y = A B C



Model 2:



Y = A B D



Model 3:



Y = A B E



Model 4:



Y = A B F



If variables D and E were significant (but not C) my list would look like:



D

E



I can handle adding the variables to the model one by one but all I have
come up with for the list creation so far is to use OMS to output to a file
that I can then open and check for significance and then output the variable
name to a file if it is significant.  I can work this out programmatically,
but the two main problems with this are that it involves a lot of opening
and creating of files (and I expect it to therefore be rather inefficient)
and I'm not sure how to collect all the variable names in one list.



I learning to use the Python plugin so I imagine that that would be one way
to go, but I'm open to other suggestions.



Any help is appreciated!



Bryan

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

[Peck, Jon]

You might get some misleading results with that many individual significance tests, but it is easy to cull the information you want without even resorting to Python.

Run syntax like this.  This example does all the regressions at once, but there could be many regression commands between OMS and OMSEND.

This will create a dataset named coef with all the coefficient statistics.  Then you can use a standard SPSS filter to select out just the significant ones based on the Sig variable in that dataset.  There will be multiple instances of a variable if it is significant more than once, but you could screen those out with Identify Duplicate Cases on the Data menu or by using AGGREGATE to pick the first/last/max/min instance.

dataset declare coef.
oms select tables /if SUBTYPES ='Coefficients'
/DESTINATION outfile=coef format=SAV.
REGRESSION
  /MISSING LISTWISE
  /STATISTICS COEFF OUTS R ANOVA
  /CRITERIA=PIN(.05) POUT(.10)
  /NOORIGIN
  /DEPENDENT accel
  /METHOD=ENTER engine
  /METHOD=ENTER horse
  /METHOD=ENTER mpg
  /METHOD=ENTER weight.
omsend.

HTH,
Jon Peck

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Bryan Tec
Sent: Friday, June 06, 2008 7:08 AM
To: [hidden email]
Subject: [SPSSX-L] Creating a list of significant variables

Hi all,



I am doing some exploratory work involving adding variables one at a time to
a base regression model.  There is a large number of variables to be added
one at a time and hence a large number of models to be run.  Rather than
manually going through them to find the significant results I would like to
have SPSS create a list of the variables that were significant.  For
example:



DV: Y

Standard IVs: A and B

List of IVs to be added one at a time: C, D, E, F.



Model 1:



Y = A B C



Model 2:



Y = A B D



Model 3:



Y = A B E



Model 4:



Y = A B F



If variables D and E were significant (but not C) my list would look like:



D

E



I can handle adding the variables to the model one by one but all I have
come up with for the list creation so far is to use OMS to output to a file
that I can then open and check for significance and then output the variable
name to a file if it is significant.  I can work this out programmatically,
but the two main problems with this are that it involves a lot of opening
and creating of files (and I expect it to therefore be rather inefficient)
and I'm not sure how to collect all the variable names in one list.



I learning to use the Python plugin so I imagine that that would be one way
to go, but I'm open to other suggestions.



Any help is appreciated!



Bryan

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

[Art Kendall]

I may not have a firm grasp of your question, but try something like this
open a new session of SPSS. Cut, paste this syntax into a syntax window.
change the GET to match your setup. Then run the syntax.

GET
  FILE='C:\Program Files (x86)\SPSSInc\SPSS16\Samples\car_sales.sav'.
DATASET NAME DataSet1 WINDOW=FRONT.REGRESSION
   variables= price engine_s horsepow wheelbas width length curb_wgt
fuel_cap mpg
  /DEPENDENT price
  /METHOD=ENTER engine_s horsepow
  /METHOD=ENTER wheelbas width length curb_wgt fuel_cap mpg.


Look at the Excluded Variables table and cut-and paste or OMS the table
into another file so you can manipulate it.

Art Kendall
Social Research consultants

Bryan Tec wrote:

> Hi all,
>
>
>
> I am doing some exploratory work involving adding variables one at a time to
> a base regression model.  There is a large number of variables to be added
> one at a time and hence a large number of models to be run.  Rather than
> manually going through them to find the significant results I would like to
> have SPSS create a list of the variables that were significant.  For
> example:
>
>
>
> DV: Y
>
> Standard IVs: A and B
>
> List of IVs to be added one at a time: C, D, E, F.
>
>
>
> Model 1:
>
>
>
> Y = A B C
>
>
>
> Model 2:
>
>
>
> Y = A B D
>
>
>
> Model 3:
>
>
>
> Y = A B E
>
>
>
> Model 4:
>
>
>
> Y = A B F
>
>
>
> If variables D and E were significant (but not C) my list would look like:
>
>
>
> D
>
> E
>
>
>
> I can handle adding the variables to the model one by one but all I have
> come up with for the list creation so far is to use OMS to output to a file
> that I can then open and check for significance and then output the variable
> name to a file if it is significant.  I can work this out programmatically,
> but the two main problems with this are that it involves a lot of opening
> and creating of files (and I expect it to therefore be rather inefficient)
> and I'm not sure how to collect all the variable names in one list.
>
>
>
> I learning to use the Python plugin so I imagine that that would be one way
> to go, but I'm open to other suggestions.
>
>
>
> Any help is appreciated!
>
>
>
> Bryan
>
> =====================
> 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