SPSSX Discussion

saving individual growth parameters

Classic

List

Threaded

6 messages Options

Matthew Pirritano

saving individual growth parameters

All listers,

I am looking for a way to save individual growth parameters that result from a mixed model. I've searched this in the archives and all I can see is that I can save the predicted values and the standard errors. But I want the slopes and the intercepts. I want to then enter these into a logistic regression where I use slopes and intercepts to predict an outcome. I cannot just run a longitudinal logistic mixed model because the dichotomous dv only occurs at the last time point, time 3.

Any advice would be much appreciated.

Thanks

Matt

Matthew Pirritano, Ph.D.
Email: [hidden email]

Maguin, Eugene

Re: saving individual growth parameters

Matt,

I don't know whether you has access to SEM programs or not but they would be the method of choice because you can fit the entire model (mixed model plus the dichotomous DV) in one instance. Example programs would be Mplus, EQS, Lisrel, MX. I assume Amos will as well but i am not sure about it. With Mplus, the mixed model part of the model is limited to two levels or three levels if the lowest level is repeated measures.

That aside, I wonder if you could extract the intercept and curve coefficients from the predicted values. I explicitly assume a growth curve model when i suggest this. If you had a linear slope model, the predicted values would fit the algebra equation for a line for each level 2 person: y=mx+b. I have not tried this so there could easily be something that i have missed or do not understand.

Gene Maguin

On Fri 12/02/11 8:39 PM , Matthew Pirritano [hidden email] sent:

> All listers,
> I am looking for a way to save individual growth parameters that
> result from a mixed model. I've searched this in the archives and all
> I can see is that I can save the predicted values and the standard
> errors. But I want the slopes and the intercepts. I want to then enter
> these into a logistic regression where I use slopes and intercepts to
> predict an outcome. I cannot just run a longitudinal logistic mixed
> model because the dichotomous dv only occurs at the last time point,
> time 3.
> Any advice would be much appreciated.
> ThanksMatt
> Matthew Pirritano, Ph.D.
> Email: [hidden email]
>
>

Matthew Pirritano

Re: saving individual growth parameters

Gene,

I believe that I have done this. I could use some confirmation of my methodology. I saved the predicted values from the growth curve model. Then I ran one OLS regression per person predicting the predicted values with time. The coefficients from the output were then entered into a new datafile using OMS. Those would be the slope and intercept per person that would then be used to predict the dichotomous outcome in a logistic regression. Does that make sense?

Thanks!
Matt

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of [hidden email]
Sent: Saturday, December 03, 2011 6:13 AM
To: [hidden email]
Subject: Re: saving individual growth parameters

Matt,

I don't know whether you has access to SEM programs or not but they would be the method of choice because you can fit the entire model (mixed model plus the dichotomous DV) in one instance. Example programs would be Mplus, EQS, Lisrel, MX. I assume Amos will as well but i am not sure about it. With Mplus, the mixed model part of the model is limited to two levels or three levels if the lowest level is repeated measures.

That aside, I wonder if you could extract the intercept and curve coefficients from the predicted values. I explicitly assume a growth curve model when i suggest this. If you had a linear slope model, the predicted values would fit the algebra equation for a line for each level 2 person: y=mx+b. I have not tried this so there could easily be something that i have missed or do not understand.

Gene Maguin

On Fri 12/02/11 8:39 PM , Matthew Pirritano [hidden email] sent:

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

Maguin, Eugene

Re: saving individual growth parameters

-----Original Message-----
From: Matthew Pirritano [mailto:[hidden email]]
Sent: Saturday, December 03, 2011 12:37 PM
To: [hidden email]; [hidden email]
Subject: RE: saving individual growth parameters

Gene,

I believe that I have done this. I could use some confirmation of my methodology. I saved the predicted values from the growth curve model. Then I ran one OLS regression per person predicting the predicted values with time. The coefficients from the output were then entered into a new datafile using OMS. Those would be the slope and intercept per person that would then be used to predict the dichotomous outcome in a logistic regression. Does that make sense?

Thanks!
Matt

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of [hidden email]
Sent: Saturday, December 03, 2011 6:13 AM
To: [hidden email]
Subject: Re: saving individual growth parameters

Matt,

I don't know whether you has access to SEM programs or not but they would be the method of choice because you can fit the entire model (mixed model plus the dichotomous DV) in one instance. Example programs would be Mplus, EQS, Lisrel, MX. I assume Amos will as well but i am not sure about it. With Mplus, the mixed model part of the model is limited to two levels or three levels if the lowest level is repeated measures.

That aside, I wonder if you could extract the intercept and curve coefficients from the predicted values. I explicitly assume a growth curve model when i suggest this. If you had a linear slope model, the predicted values would fit the algebra equation for a line for each level 2 person: y=mx+b. I have not tried this so there could easily be something that i have missed or do not understand.

Gene Maguin

On Fri 12/02/11 8:39 PM , Matthew Pirritano [hidden email] sent:

Maguin, Eugene

Re: saving individual growth parameters

In reply to this post by Matthew Pirritano

Sorry for the preceding empty message.

Matt,
Yes, that is exactly what I had in mind. I didn't realize that you had already done the within person regressions.

Gene

-----Original Message-----
From: Matthew Pirritano [mailto:[hidden email]]
Sent: Saturday, December 03, 2011 12:37 PM
To: [hidden email]; [hidden email]
Subject: RE: saving individual growth parameters

Gene,

I believe that I have done this. I could use some confirmation of my methodology. I saved the predicted values from the growth curve model. Then I ran one OLS regression per person predicting the predicted values with time. The coefficients from the output were then entered into a new datafile using OMS. Those would be the slope and intercept per person that would then be used to predict the dichotomous outcome in a logistic regression. Does that make sense?

Thanks!
Matt

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of [hidden email]
Sent: Saturday, December 03, 2011 6:13 AM
To: [hidden email]
Subject: Re: saving individual growth parameters

Matt,

I don't know whether you has access to SEM programs or not but they would be the method of choice because you can fit the entire model (mixed model plus the dichotomous DV) in one instance. Example programs would be Mplus, EQS, Lisrel, MX. I assume Amos will as well but i am not sure about it. With Mplus, the mixed model part of the model is limited to two levels or three levels if the lowest level is repeated measures.

That aside, I wonder if you could extract the intercept and curve coefficients from the predicted values. I explicitly assume a growth curve model when i suggest this. If you had a linear slope model, the predicted values would fit the algebra equation for a line for each level 2 person: y=mx+b. I have not tried this so there could easily be something that i have missed or do not understand.

Gene Maguin

On Fri 12/02/11 8:39 PM , Matthew Pirritano [hidden email] sent:

Matthew Pirritano

Re: saving individual growth parameters

Gene,

Thanks for confirming my notion!

Thanks
Matt

-----Original Message-----
From: Gene Maguin [mailto:[hidden email]]
Sent: Sunday, December 04, 2011 10:28 AM
To: 'Matthew Pirritano'; [hidden email]
Subject: RE: saving individual growth parameters

Sorry for the preceding empty message.

Matt,
Yes, that is exactly what I had in mind. I didn't realize that you had already done the within person regressions.

Gene

-----Original Message-----
From: Matthew Pirritano [mailto:[hidden email]]
Sent: Saturday, December 03, 2011 12:37 PM
To: [hidden email]; [hidden email]
Subject: RE: saving individual growth parameters

Gene,

I believe that I have done this. I could use some confirmation of my methodology. I saved the predicted values from the growth curve model. Then I ran one OLS regression per person predicting the predicted values with time. The coefficients from the output were then entered into a new datafile using OMS. Those would be the slope and intercept per person that would then be used to predict the dichotomous outcome in a logistic regression. Does that make sense?

Thanks!
Matt

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of [hidden email]
Sent: Saturday, December 03, 2011 6:13 AM
To: [hidden email]
Subject: Re: saving individual growth parameters

Matt,

I don't know whether you has access to SEM programs or not but they would be the method of choice because you can fit the entire model (mixed model plus the dichotomous DV) in one instance. Example programs would be Mplus, EQS, Lisrel, MX. I assume Amos will as well but i am not sure about it. With Mplus, the mixed model part of the model is limited to two levels or three levels if the lowest level is repeated measures.

That aside, I wonder if you could extract the intercept and curve coefficients from the predicted values. I explicitly assume a growth curve model when i suggest this. If you had a linear slope model, the predicted values would fit the algebra equation for a line for each level 2 person: y=mx+b. I have not tried this so there could easily be something that i have missed or do not understand.

Gene Maguin

On Fri 12/02/11 8:39 PM , Matthew Pirritano [hidden email] sent: