compute probabilities (discriminant analysis)

[drfg2008]

sorry for a somewhat fundamental question:

Does anyone know the formula how to compute the (group-) probabilities from the discriminant scores? Any internet accessible literature is welcome.

(most descriptions of the discriminant analysis describe how to get the discriminant scores, but amazingly do not provide any clue how to compute the probabilities)

Thanks

[Art Kendall]

If by "group probabilities" you mean the probabilities for each case that are added to the file with the /save option that should be in the algorithms help.

If by "group probabilities" , you mean the crosstab of original by assigned groups from the classification phase that should also be in the algorithms help.

If these are not what you mean please give more detail.

Art Kendall
Social Research Consultants

On 11/14/2011 4:26 AM, drfg2008 wrote:

sorry for a somewhat fundamental question:

Does anyone know the formula how to compute the (group-) probabilities from
the discriminant scores? Any internet accessible literature is welcome.

(most descriptions of the discriminant analysis describe how to get the
discriminant scores, but amazingly do not provide any clue how to compute
the probabilities)

Thanks



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

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/compute-probabilities-discriminant-analysis-tp4990219p4990219.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

[drfg2008]

Thanks,

I meant the probabilities for each case that are added to the file with the /save option

I've found it at last. Seems to be quite a challenge for the programming department. Isn't there a more simple way for programmers, something like an approximation ?

Frank

[Art Kendall]

There are two ways to apply the model to other cases.
If you are on a server version you can use OUTFILE to produce XML.

Or you can do it the old fashioned way.
keep the original data and syntax.
when you have new cases for which you do not know the group, give them a value that is outside the range of values for the original groups.
Add cases.
Run the syntax.
the ungrouped cases will have assigned groups and probabilities for each of the original groups.

Art Kendall
Social Research Consultants

On 11/14/2011 8:18 AM, drfg2008 wrote:

Thanks,

I meant the probabilities for each case that are added to the file with the
/save option

I've found it at last. Seems to be quite a challenge for the programming
department. Isn't there a more simple way for programmers, something like an
approximation ?

Frank



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

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/compute-probabilities-discriminant-analysis-tp4990219p4990733.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

[Jon K Peck]

Remember that Statistics now has a scoring wizard that can score lots of models.  This is included in the Client as well as the Server version.

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




From:        Art Kendall <[hidden email]>
To:        [hidden email]
Date:        11/14/2011 09:09 AM
Subject:        Re: [SPSSX-L] compute probabilities (discriminant analysis)
Sent by:        "SPSSX(r) Discussion" <[hidden email]>

---

There are two ways to apply the model to other cases.
If you are on a server version you can use OUTFILE to produce XML.

Or you can do it the old fashioned way.
keep the original data and syntax.
when you have new cases for which you do not know the group, give them a value that is outside the range of values for the original groups.
Add cases.
Run the syntax.
the ungrouped cases will have assigned groups and probabilities for each of the original groups.

Art Kendall
Social Research Consultants

On 11/14/2011 8:18 AM, drfg2008 wrote:
Thanks,

I meant the probabilities for each case that are added to the file with the
/save option

I've found it at last. Seems to be quite a challenge for the programming
department. Isn't there a more simple way for programmers, something like an
approximation ?

Frank



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

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/compute-probabilities-discriminant-analysis-tp4990219p4990733.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@... (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]

Glad to know that the wizard is available in the client version.  Is that new or did I remember incorrectly that it worked only when there was a serve?

Art

On 11/14/2011 11:51 AM, Jon K Peck wrote:

Remember that Statistics now has a scoring wizard that can score lots of models.  This is included in the Client as well as the Server version.

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




From:        Art Kendall [hidden email]
To:        [hidden email]
Date:        11/14/2011 09:09 AM
Subject:        Re: [SPSSX-L] compute probabilities (discriminant analysis)
Sent by:        "SPSSX(r) Discussion" [hidden email]

---

There are two ways to apply the model to other cases.
If you are on a server version you can use OUTFILE to produce XML.

Or you can do it the old fashioned way.
keep the original data and syntax.
when you have new cases for which you do not know the group, give them a value that is outside the range of values for the original groups.
Add cases.
Run the syntax.
the ungrouped cases will have assigned groups and probabilities for each of the original groups.

Art Kendall
Social Research Consultants

On 11/14/2011 8:18 AM, drfg2008 wrote:
Thanks,

I meant the probabilities for each case that are added to the file with the
/save option

I've found it at last. Seems to be quite a challenge for the programming
department. Isn't there a more simple way for programmers, something like an
approximation ?

Frank



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

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/compute-probabilities-discriminant-analysis-tp4990219p4990733.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

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

[Jon K Peck]

The original scoring functionality was Server only (via the ApplyModel function).  The scoring wizard was new in V19 IIRC but was included in the Client version.

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




From:        Art Kendall <[hidden email]>
To:        Jon K Peck/Chicago/IBM@IBMUS
Cc:        [hidden email]
Date:        11/14/2011 10:56 AM
Subject:        Re: [SPSSX-L] compute probabilities (discriminant analysis)

---

Glad to know that the wizard is available in the client version.  Is that new or did I remember incorrectly that it worked only when there was a serve?

Art

On 11/14/2011 11:51 AM, Jon K Peck wrote:
Remember that Statistics now has a scoring wizard that can score lots of models.  This is included in the Client as well as the Server version.

Jon Peck (no "h")
Senior Software Engineer, IBM
peck@...
new phone: [phone]




From:        Art Kendall <Art@...>
To:        [hidden email]
Date:        11/14/2011 09:09 AM
Subject:        Re: [SPSSX-L] compute probabilities (discriminant analysis)
Sent by:        "SPSSX(r) Discussion" [hidden email]

---

There are two ways to apply the model to other cases.
If you are on a server version you can use OUTFILE to produce XML.

Or you can do it the old fashioned way.
keep the original data and syntax.
when you have new cases for which you do not know the group, give them a value that is outside the range of values for the original groups.
Add cases.
Run the syntax.
the ungrouped cases will have assigned groups and probabilities for each of the original groups.

Art Kendall
Social Research Consultants

On 11/14/2011 8:18 AM, drfg2008 wrote:
Thanks,

I meant the probabilities for each case that are added to the file with the
/save option

I've found it at last. Seems to be quite a challenge for the programming
department. Isn't there a more simple way for programmers, something like an
approximation ?

Frank



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

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/compute-probabilities-discriminant-analysis-tp4990219p4990733.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
LISTSERV@... (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 LISTSERV@... (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