SPSSX Discussion

Grouping logistic probabilites

Classic

List

Threaded

7 messages Options

Jignesh Sutar

Grouping logistic probabilites

Hello,

I am running a logistic regression model which saves the predicted
probabilities for those who did and did not answer on the dependent.

I would like to group the probabilities into 5 tiles based on those who were
modelled. I would like to classify the non-responders into the same groups
depending on their predicted probability score.

I have tried RANK command but this doesn't quite work for my purpose.

Many thanks in advance.

ATB,
Jignesh

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

ViAnn Beadle

Re: Grouping logistic probabilites

In what way does RANK not work for you?

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Sutar
Sent: Thursday, May 01, 2008 6:30 AM
To: [hidden email]
Subject: Grouping logistic probabilites

Hello,

I am running a logistic regression model which saves the predicted
probabilities for those who did and did not answer on the dependent.

I would like to group the probabilities into 5 tiles based on those who were
modelled. I would like to classify the non-responders into the same groups
depending on their predicted probability score.

I have tried RANK command but this doesn't quite work for my purpose.

Many thanks in advance.

ATB,
Jignesh

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

Jignesh Sutar

Re: Grouping logistic probabilites

My current attempt is:

RANK VARIABLES= PRE_1 (A) BY DummyV1
/ntiles(5) INTO RANK1.

where DummyV1 =1 if valid responses on DV
DummyV1=0 if missing at DV

I would like to group PRE_1 (the predicted probabilites) into 5 equal groups
for where DummyV1=1 only. So the break down of these 5 groups will be based
on the probability score.

I would then like to classify those who belong to DummyV1=0 into the same
definition of the 5 groups already created for DummyV1=1.

Currently my syntax ranks both groups on DummyV1 into 5 equal sized groups
individually.

On 01/05/2008, ViAnn Beadle <[hidden email]> wrote:

>
> In what way does RANK not work for you?
>
> -----Original Message-----
> From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
> Sutar
> Sent: Thursday, May 01, 2008 6:30 AM
> To: [hidden email]
> Subject: Grouping logistic probabilites
>
> Hello,
>
> I am running a logistic regression model which saves the predicted
> probabilities for those who did and did not answer on the dependent.
>
> I would like to group the probabilities into 5 tiles based on those who
> were
> modelled. I would like to classify the non-responders into the same groups
> depending on their predicted probability score.
>
> I have tried RANK command but this doesn't quite work for my purpose.
>
> Many thanks in advance.
>
> ATB,
> Jignesh
>
> =====================
> 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

Jignesh Sutar

Re: Grouping logistic probabilites

Here is a little bit more information,

This is a similar example of what I require but this example involves manual
intervention in the form of copying and pasting output tables from SPSS to
excel to produce the IF statements below. I am hoping I can learn how to
this more efficiently.

USE ALL.
COMPUTE filter_$=(DummyV1=1).
VARIABLE LABEL filter_$ 'DummyV1=1 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE .

FREQUENCIES
VARIABLES=PRE_1
/NTILES= 10
/ORDER= ANALYSIS .

IF (PRE_1 <=0.00164676643061749) RankGrp=1.
IF (PRE_1 >0.00164676643061749 & PRE_1 <= 0.00412623256883683) RankGrp=2.
IF (PRE_1 >0.00412623256883683 & PRE_1 <= 0.00733462954248551) RankGrp=3.
IF (PRE_1 >0.00733462954248551 & PRE_1 <= 0.0109616626838707) RankGrp=4.
IF (PRE_1 >0.0109616626838707 & PRE_1 <= 0.0156787765386074) RankGrp=5.
IF (PRE_1 >0.0156787765386074 & PRE_1 <= 0.0230233936216253) RankGrp=6.
IF (PRE_1 >0.0230233936216253 & PRE_1 <= 0.0345075335250341) RankGrp=7.
IF (PRE_1 >0.0345075335250341 & PRE_1 <= 0.0572568186083756) RankGrp=8.
IF (PRE_1 >0.0572568186083756 & PRE_1 <= 0.139969182373795) RankGrp=9.
IF (PRE_1 >0.139969182373795) RankGrp=10.

On 01/05/2008, J Sutar <[hidden email]> wrote:

>
> My current attempt is:
>
> RANK VARIABLES= PRE_1 (A) BY DummyV1
> /ntiles(5) INTO RANK1.
>
>
> where DummyV1 =1 if valid responses on DV
> DummyV1=0 if missing at DV
>
> I would like to group PRE_1 (the predicted probabilites) into 5 equal
> groups for where DummyV1=1 only. So the break down of these 5 groups will be
> based on the probability score.
>
> I would then like to classify those who belong to DummyV1=0 into the same
> definition of the 5 groups already created for DummyV1=1.
>
>
>
> Currently my syntax ranks both groups on DummyV1 into 5 equal sized groups
> individually.
>
>
> On 01/05/2008, ViAnn Beadle <[hidden email]> wrote:
> >
> > In what way does RANK not work for you?
> >
> > -----Original Message-----
> > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
> > J
> > Sutar
> > Sent: Thursday, May 01, 2008 6:30 AM
> > To: [hidden email]
> > Subject: Grouping logistic probabilites
> >
> > Hello,
> >
> > I am running a logistic regression model which saves the predicted
> > probabilities for those who did and did not answer on the dependent.
> >
> > I would like to group the probabilities into 5 tiles based on those who
> > were
> > modelled. I would like to classify the non-responders into the same
> > groups
> > depending on their predicted probability score.
> >
> > I have tried RANK command but this doesn't quite work for my purpose.
> >
> > Many thanks in advance.
> >
> > ATB,
> > Jignesh
> >
> > =====================
> > 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
> >
> >
>

Jignesh Sutar

Re: Grouping logistic probabilites

Hi,

I was wondering if anyone was able to help with the below please?

Thanks
Jignesh

2008/5/1 J Sutar <[hidden email]>:

> Here is a little bit more information,
>
> This is a similar example of what I require but this example involves
> manual intervention in the form of copying and pasting output tables from
> SPSS to excel to produce the IF statements below. I am hoping I can learn
> how to this more efficiently.
>
>
> USE ALL.
> COMPUTE filter_$=(DummyV1=1).
> VARIABLE LABEL filter_$ 'DummyV1=1 (FILTER)'.
> VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
> FORMAT filter_$ (f1.0).
> FILTER BY filter_$.
> EXECUTE .
>
>
> FREQUENCIES
> VARIABLES=PRE_1
> /NTILES= 10
> /ORDER= ANALYSIS .
>
> IF (PRE_1 <=0.00164676643061749) RankGrp=1.
> IF (PRE_1 >0.00164676643061749 & PRE_1 <= 0.00412623256883683) RankGrp=2.
> IF (PRE_1 >0.00412623256883683 & PRE_1 <= 0.00733462954248551) RankGrp=3.
> IF (PRE_1 >0.00733462954248551 & PRE_1 <= 0.0109616626838707) RankGrp=4.
> IF (PRE_1 >0.0109616626838707 & PRE_1 <= 0.0156787765386074) RankGrp=5.
> IF (PRE_1 >0.0156787765386074 & PRE_1 <= 0.0230233936216253) RankGrp=6.
> IF (PRE_1 >0.0230233936216253 & PRE_1 <= 0.0345075335250341) RankGrp=7.
> IF (PRE_1 >0.0345075335250341 & PRE_1 <= 0.0572568186083756) RankGrp=8.
> IF (PRE_1 >0.0572568186083756 & PRE_1 <= 0.139969182373795) RankGrp=9.
> IF (PRE_1 >0.139969182373795) RankGrp=10.
>
>
>
> On 01/05/2008, J Sutar <[hidden email]> wrote:
> >
> > My current attempt is:
> >
> > RANK VARIABLES= PRE_1 (A) BY DummyV1
> > /ntiles(5) INTO RANK1.
> >
> >
> > where DummyV1 =1 if valid responses on DV
> > DummyV1=0 if missing at DV
> >
> > I would like to group PRE_1 (the predicted probabilites) into 5 equal
> > groups for where DummyV1=1 only. So the break down of these 5 groups will be
> > based on the probability score.
> >
> > I would then like to classify those who belong to DummyV1=0 into the
> > same definition of the 5 groups already created for DummyV1=1.
> >
> >
> >
> > Currently my syntax ranks both groups on DummyV1 into 5 equal sized
> > groups individually.
> >
> >
> > On 01/05/2008, ViAnn Beadle <[hidden email]> wrote:
> > >
> > > In what way does RANK not work for you?
> > >
> > > -----Original Message-----
> > > From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf
> > > Of J
> > > Sutar
> > > Sent: Thursday, May 01, 2008 6:30 AM
> > > To: [hidden email]
> > > Subject: Grouping logistic probabilites
> > >
> > > Hello,
> > >
> > > I am running a logistic regression model which saves the predicted
> > > probabilities for those who did and did not answer on the dependent.
> > >
> > > I would like to group the probabilities into 5 tiles based on those
> > > who were
> > > modelled. I would like to classify the non-responders into the same
> > > groups
> > > depending on their predicted probability score.
> > >
> > > I have tried RANK command but this doesn't quite work for my purpose.
> > >
> > > Many thanks in advance.
> > >
> > > ATB,
> > > Jignesh
> > >
> > > =====================
> > > 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
> > >
> > >
> >
>

Albert-Jan Roskam

Re: Grouping logistic probabilites

Hi,

I'm not sure if you want quintiles or deciles. It
seems that you were very close to the solution
already. The following code is untested:

compute myfilter = (dummyv1=1).
filter by myfilter.
rank variables pre_1
/ntiles (10) into pre_1_decile
/ntiles (5) into pre_1_quintile.
filter off.

Cheers!!
Albert-Jan

--- J Sutar <[hidden email]> wrote:

> Hi,
>
> I was wondering if anyone was able to help with the
> below please?
>
> Thanks
> Jignesh
>
> 2008/5/1 J Sutar <[hidden email]>:
>
> > Here is a little bit more information,
> >
> > This is a similar example of what I require but
> this example involves
> > manual intervention in the form of copying and
> pasting output tables from
> > SPSS to excel to produce the IF statements below.
> I am hoping I can learn
> > how to this more efficiently.
> >
> >
> > USE ALL.
> > COMPUTE filter_$=(DummyV1=1).
> > VARIABLE LABEL filter_$ 'DummyV1=1 (FILTER)'.
> > VALUE LABELS filter_$ 0 'Not Selected' 1
> 'Selected'.
> > FORMAT filter_$ (f1.0).
> > FILTER BY filter_$.
> > EXECUTE .
> >
> >
> > FREQUENCIES
> > VARIABLES=PRE_1
> > /NTILES= 10
> > /ORDER= ANALYSIS .
> >
> > IF (PRE_1 <=0.00164676643061749) RankGrp=1.
> > IF (PRE_1 >0.00164676643061749 & PRE_1 <=
> 0.00412623256883683) RankGrp=2.
> > IF (PRE_1 >0.00412623256883683 & PRE_1 <=
> 0.00733462954248551) RankGrp=3.
> > IF (PRE_1 >0.00733462954248551 & PRE_1 <=
> 0.0109616626838707) RankGrp=4.
> > IF (PRE_1 >0.0109616626838707 & PRE_1 <=
> 0.0156787765386074) RankGrp=5.
> > IF (PRE_1 >0.0156787765386074 & PRE_1 <=
> 0.0230233936216253) RankGrp=6.
> > IF (PRE_1 >0.0230233936216253 & PRE_1 <=
> 0.0345075335250341) RankGrp=7.
> > IF (PRE_1 >0.0345075335250341 & PRE_1 <=
> 0.0572568186083756) RankGrp=8.
> > IF (PRE_1 >0.0572568186083756 & PRE_1 <=
> 0.139969182373795) RankGrp=9.
> > IF (PRE_1 >0.139969182373795) RankGrp=10.
> >
> >
> >
> > On 01/05/2008, J Sutar <[hidden email]> wrote:
> > >
> > > My current attempt is:
> > >
> > > RANK VARIABLES= PRE_1 (A) BY DummyV1
> > > /ntiles(5) INTO RANK1.
> > >
> > >
> > > where DummyV1 =1 if valid responses on DV
> > > DummyV1=0 if missing at DV
> > >
> > > I would like to group PRE_1 (the predicted
> probabilites) into 5 equal
> > > groups for where DummyV1=1 only. So the break
> down of these 5 groups will be
> > > based on the probability score.
> > >
> > > I would then like to classify those who belong
> to DummyV1=0 into the
> > > same definition of the 5 groups already created
> for DummyV1=1.
> > >
> > >
> > >
> > > Currently my syntax ranks both groups on DummyV1
> into 5 equal sized
> > > groups individually.
> > >
> > >
> > > On 01/05/2008, ViAnn Beadle <[hidden email]>
> wrote:
> > > >
> > > > In what way does RANK not work for you?
> > > >
> > > > -----Original Message-----
> > > > From: SPSSX(r) Discussion
> [mailto:[hidden email]] On Behalf
> > > > Of J
> > > > Sutar
> > > > Sent: Thursday, May 01, 2008 6:30 AM
> > > > To: [hidden email]
> > > > Subject: Grouping logistic probabilites
> > > >
> > > > Hello,
> > > >
> > > > I am running a logistic regression model which
> saves the predicted
> > > > probabilities for those who did and did not
> answer on the dependent.
> > > >
> > > > I would like to group the probabilities into 5
> tiles based on those
> > > > who were
> > > > modelled. I would like to classify the
> non-responders into the same
> > > > groups
> > > > depending on their predicted probability
> score.
> > > >
> > > > I have tried RANK command but this doesn't
> quite work for my purpose.
> > > >
> > > > Many thanks in advance.
> > > >
> > > > ATB,
> > > > Jignesh
> > > >
> > > > =====================
> > > > 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
>

____________________________________________________________________________________
Be a better friend, newshound, and
know-it-all with Yahoo! Mobile. Try it now. http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ

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

Jignesh Sutar

Re: Grouping logistic probabilites

Hi,

Yes almost, the syntax you provided me does the same as I have so far. Which
is to break for dummyv1=1 into deciles. A further step i would like to
compute is to break for dummyv1=0 into 'deciles' (unequal sized) also,
defined by the cutpoints of dummyv1=1.

Does that makes sense?

Is this doable?

Jignesh

On 03/05/2008, Albert-jan Roskam <[hidden email]> wrote:

>
> Hi,
>
> I'm not sure if you want quintiles or deciles. It
> seems that you were very close to the solution
> already. The following code is untested:
>
> compute myfilter = (dummyv1=1).
> filter by myfilter.
> rank variables pre_1
> /ntiles (10) into pre_1_decile
> /ntiles (5) into pre_1_quintile.
> filter off.
>
> Cheers!!
> Albert-Jan
>
> --- J Sutar <[hidden email]> wrote:
>
> > Hi,
> >
> > I was wondering if anyone was able to help with the
> > below please?
> >
> > Thanks
> > Jignesh
> >
> > 2008/5/1 J Sutar <[hidden email]>:
> >
> > > Here is a little bit more information,
> > >
> > > This is a similar example of what I require but
> > this example involves
> > > manual intervention in the form of copying and
> > pasting output tables from
> > > SPSS to excel to produce the IF statements below.
> > I am hoping I can learn
> > > how to this more efficiently.
> > >
> > >
> > > USE ALL.
> > > COMPUTE filter_$=(DummyV1=1).
> > > VARIABLE LABEL filter_$ 'DummyV1=1 (FILTER)'.
> > > VALUE LABELS filter_$ 0 'Not Selected' 1
> > 'Selected'.
> > > FORMAT filter_$ (f1.0).
> > > FILTER BY filter_$.
> > > EXECUTE .
> > >
> > >
> > > FREQUENCIES
> > > VARIABLES=PRE_1
> > > /NTILES= 10
> > > /ORDER= ANALYSIS .
> > >
> > > IF (PRE_1 <=0.00164676643061749) RankGrp=1.
> > > IF (PRE_1 >0.00164676643061749 & PRE_1 <=
> > 0.00412623256883683) RankGrp=2.
> > > IF (PRE_1 >0.00412623256883683 & PRE_1 <=
> > 0.00733462954248551) RankGrp=3.
> > > IF (PRE_1 >0.00733462954248551 & PRE_1 <=
> > 0.0109616626838707) RankGrp=4.
> > > IF (PRE_1 >0.0109616626838707 & PRE_1 <=
> > 0.0156787765386074) RankGrp=5.
> > > IF (PRE_1 >0.0156787765386074 & PRE_1 <=
> > 0.0230233936216253) RankGrp=6.
> > > IF (PRE_1 >0.0230233936216253 & PRE_1 <=
> > 0.0345075335250341) RankGrp=7.
> > > IF (PRE_1 >0.0345075335250341 & PRE_1 <=
> > 0.0572568186083756) RankGrp=8.
> > > IF (PRE_1 >0.0572568186083756 & PRE_1 <=
> > 0.139969182373795) RankGrp=9.
> > > IF (PRE_1 >0.139969182373795) RankGrp=10.
> > >
> > >
> > >
> > > On 01/05/2008, J Sutar <[hidden email]> wrote:
> > > >
> > > > My current attempt is:
> > > >
> > > > RANK VARIABLES= PRE_1 (A) BY DummyV1
> > > > /ntiles(5) INTO RANK1.
> > > >
> > > >
> > > > where DummyV1 =1 if valid responses on DV
> > > > DummyV1=0 if missing at DV
> > > >
> > > > I would like to group PRE_1 (the predicted
> > probabilites) into 5 equal
> > > > groups for where DummyV1=1 only. So the break
> > down of these 5 groups will be
> > > > based on the probability score.
> > > >
> > > > I would then like to classify those who belong
> > to DummyV1=0 into the
> > > > same definition of the 5 groups already created
> > for DummyV1=1.
> > > >
> > > >
> > > >
> > > > Currently my syntax ranks both groups on DummyV1
> > into 5 equal sized
> > > > groups individually.
> > > >
> > > >
> > > > On 01/05/2008, ViAnn Beadle <[hidden email]>
> > wrote:
> > > > >
> > > > > In what way does RANK not work for you?
> > > > >
> > > > > -----Original Message-----
> > > > > From: SPSSX(r) Discussion
> > [mailto:[hidden email]] On Behalf
> > > > > Of J
> > > > > Sutar
> > > > > Sent: Thursday, May 01, 2008 6:30 AM
> > > > > To: [hidden email]
> > > > > Subject: Grouping logistic probabilites
> > > > >
> > > > > Hello,
> > > > >
> > > > > I am running a logistic regression model which
> > saves the predicted
> > > > > probabilities for those who did and did not
> > answer on the dependent.
> > > > >
> > > > > I would like to group the probabilities into 5
> > tiles based on those
> > > > > who were
> > > > > modelled. I would like to classify the
> > non-responders into the same
> > > > > groups
> > > > > depending on their predicted probability
> > score.
> > > > >
> > > > > I have tried RANK command but this doesn't
> > quite work for my purpose.
> > > > >
> > > > > Many thanks in advance.
> > > > >
> > > > > ATB,
> > > > > Jignesh
> > > > >
> > > > > =====================
> > > > > 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
> >
>
>
>
>
> ____________________________________________________________________________________
> Be a better friend, newshound, and
> know-it-all with Yahoo! Mobile. Try it now.
> http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
>