Multinomial vs. Ordered logistic regression

[nina]

Dear all,

I am not sure whether the intercepts estimated in SPSS using then multinomial-regression routine are comparable (= equivalent) to the thresholds estimated in SPSS using the ordered logistic regression routine. Ultmately, my dependent variable is categorical, so I'd prefer the ordered logistic regression approach. However, I can't see why the intercepts/cut off values should be different...could anybody help me in this regard?

Many thanks,
Nina

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

[Ryan]

Nina,

Here are the equations underlying both models:

https://listserv.uga.edu/cgi-bin/wa?A2=ind1402&L=SPSSX-L&P=R40725

Unless your categorical DV has ordinal properties I see no reason you would want to employ ordinal logistic regression.

Ryan

Sent from my iPhone

> On Jan 3, 2015, at 10:08 AM, Nina Lasek <[hidden email]> wrote:
>
> Dear all,
>
> I am not sure whether the intercepts estimated in SPSS using then multinomial-regression routine are comparable (= equivalent) to the thresholds estimated in SPSS using the ordered logistic regression routine. Ultmately, my dependent variable is categorical, so I'd prefer the ordered logistic regression approach. However, I can't see why the intercepts/cut off values should be different...could anybody help me in this regard?
>
> Many thanks,
> Nina
>
> =====================
> 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

[Ryan]

Dear SPSS,

I've been spending sometime evaluating random effects ordinal logistic regression recently and it dawned on me that my previous message to this SPSS thread provided underlying equations which were not consistent with the SPSS PLUM procedure.

Please see the link I sent a while back showing the equations consistent with PLUM:

https://listserv.uga.edu/cgi-bin/wa?A2=ind1210&L=SPSSX-L&P=R89892

I actually then went on to show a numerical example here:

https://listserv.uga.edu/cgi-bin/wa?A2=ind1210&L=SPSSX-L&P=R95755

Sorry.

Ryan

On Sat, Jan 3, 2015 at 2:28 PM, <[hidden email]> wrote:

Nina,

Here are the equations underlying both models:

https://listserv.uga.edu/cgi-bin/wa?A2=ind1402&L=SPSSX-L&P=R40725

Unless your categorical DV has ordinal properties I see no reason you would want to employ ordinal logistic regression.

Ryan

Sent from my iPhone

> On Jan 3, 2015, at 10:08 AM, Nina Lasek <[hidden email]> wrote:
>
> Dear all,
>
> I am not sure whether the intercepts estimated in SPSS using then multinomial-regression routine are comparable (= equivalent) to the thresholds estimated in SPSS using the ordered logistic regression routine. Ultmately, my dependent variable is categorical, so I'd prefer the ordered logistic regression approach. However, I can't see why the intercepts/cut off values should be different...could anybody help me in this regard?
>
> Many thanks,
> Nina
>
> =====================
> 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