Interpretation of parameters in loglinear models

[Abigail]

Hello,

I am trouble using the parameter estimates table to interpret the effects in my loglinear model-paremeter.estimates.xls

Previously I have done crosstabs to examine 2-way terms in my models but now I have one variable which is a continous covariate (forearm) so now I am looking at the parameter estimates to interpret the terms that include my covariate: posture*forearm and diam*no.of.sup*forearm.

I know that SPSS takes the last category in the variable as a baseline and therefore does not provide parameter estimates for them. Within my variables there are no clear base-line categories, I really need to get values for all of the category combinations in order to fully interpret the terms in the model but I do not know how to work these out. For the 3-way relationship it is even more difficult because it only provides parameter estimates for 1 of the 4 combinations.

Does anyone know how to work out the values that have been set to zero? Or is there a better way to interpret interaction terms that involve a covariate?

Thanks,

Abigail

[Bruce Weaver]

I would work out fitted values of Y for various interesting combinations of the interacting variables, and plot them to display visually the nature of the interaction.  This is usually much easier than trying to interpret coefficients for product terms.  (Although interpretation of those coefficients becomes a lot easier when you're looking at such a plot, I find.)

Abigail wrote

Hello,

I am trouble using the parameter estimates table to interpret the effects in my loglinear model-paremeter.estimates.xls

Previously I have done crosstabs to examine 2-way terms in my models but now I have one variable which is a continous covariate (forearm) so now I am looking at the parameter estimates to interpret the terms that include my covariate: posture*forearm and diam*no.of.sup*forearm.

I know that SPSS takes the last category in the variable as a baseline and therefore does not provide parameter estimates for them. Within my variables there are no clear base-line categories, I really need to get values for all of the category combinations in order to fully interpret the terms in the model but I do not know how to work these out. For the 3-way relationship it is even more difficult because it only provides parameter estimates for 1 of the 4 combinations.

Does anyone know how to work out the values that have been set to zero? Or is there a better way to interpret interaction terms that involve a covariate?

Thanks,

Abigail

[Abigail]

Please could you explain further what you mean by the fitted values of Y? Do you mean the expected values that can be obtained from the contingency table? But these are not available for the interactions that invlove the covariate.

Bruce Weaver wrote

I would work out fitted values of Y for various interesting combinations of the interacting variables, and plot them to display visually the nature of the interaction.  This is usually much easier than trying to interpret coefficients for product terms.  (Although interpretation of those coefficients becomes a lot easier when you're looking at such a plot, I find.)

Abigail wrote

Hello,

I am trouble using the parameter estimates table to interpret the effects in my loglinear model-paremeter.estimates.xls

Previously I have done crosstabs to examine 2-way terms in my models but now I have one variable which is a continous covariate (forearm) so now I am looking at the parameter estimates to interpret the terms that include my covariate: posture*forearm and diam*no.of.sup*forearm.

I know that SPSS takes the last category in the variable as a baseline and therefore does not provide parameter estimates for them. Within my variables there are no clear base-line categories, I really need to get values for all of the category combinations in order to fully interpret the terms in the model but I do not know how to work these out. For the 3-way relationship it is even more difficult because it only provides parameter estimates for 1 of the 4 combinations.

Does anyone know how to work out the values that have been set to zero? Or is there a better way to interpret interaction terms that involve a covariate?

Thanks,

Abigail

[Bruce Weaver]

The output you provided was a table of regression coefficients (for a Poisson regression, according to one of the footnotes).  By plugging different values of the variables into that regression equation, you can get fitted values of Y at selected combinations.  Or depending on what procedure you're using, you could get them by saving fitted values to the data file.  What procedure did you use to run the model?  It might help if you pasted the syntax.

Abigail wrote

Please could you explain further what you mean by the fitted values of Y? Do you mean the expected values that can be obtained from the contingency table? But these are not available for the interactions that invlove the covariate.

Bruce Weaver wrote

I would work out fitted values of Y for various interesting combinations of the interacting variables, and plot them to display visually the nature of the interaction.  This is usually much easier than trying to interpret coefficients for product terms.  (Although interpretation of those coefficients becomes a lot easier when you're looking at such a plot, I find.)

Abigail wrote

Hello,

I am trouble using the parameter estimates table to interpret the effects in my loglinear model-paremeter.estimates.xls

Previously I have done crosstabs to examine 2-way terms in my models but now I have one variable which is a continous covariate (forearm) so now I am looking at the parameter estimates to interpret the terms that include my covariate: posture*forearm and diam*no.of.sup*forearm.

I know that SPSS takes the last category in the variable as a baseline and therefore does not provide parameter estimates for them. Within my variables there are no clear base-line categories, I really need to get values for all of the category combinations in order to fully interpret the terms in the model but I do not know how to work these out. For the 3-way relationship it is even more difficult because it only provides parameter estimates for 1 of the 4 combinations.

Does anyone know how to work out the values that have been set to zero? Or is there a better way to interpret interaction terms that involve a covariate?

Thanks,

Abigail

[Ryan]

Abigail,

This would be a good time for you to learn more about the mathematics
behind the regression model you've specified. Understanding the
equation will help you arrive at the answer. Take a look at this
message I posted a while back. This same *logic* can be applied to
your sitation:

http://www.listserv.uga.edu/cgi-bin/wa?A2=ind1008&L=spssx-l&P=R50116

HTH,

Ryan

On Wed, Jun 1, 2011 at 11:08 AM, Abigail <[hidden email]> wrote:

> Hello,
>
> I am trouble using the parameter estimates table to interpret the effects in
> my loglinear model-
> http://spssx-discussion.1045642.n5.nabble.com/file/n4445521/paremeter.estimates.xls
> paremeter.estimates.xls
>
> Previously I have done crosstabs to examine 2-way terms in my models but now
> I have one variable which is a continous covariate (forearm) so now I am
> looking at the parameter estimates to interpret the terms that include my
> covariate: posture*forearm and diam*no.of.sup*forearm.
>
> I know that SPSS takes the last category in the variable as a baseline and
> therefore does not provide parameter estimates for them. Within my variables
> there are no clear base-line categories, I really need to get values for all
> of the category combinations in order to fully interpret the terms in the
> model but I do not know how to work these out. For the 3-way relationship it
> is even more difficult because it only provides parameter estimates for 1 of
> the 4 combinations.
>
> Does anyone know how to work out the values that have been set to zero? Or
> is there a better way to interpret interaction terms that involve a
> covariate?
>
> Thanks,
>
> Abigail
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Interpretation-of-parameters-in-loglinear-models-tp4445521p4445521.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

[Kelly Vander Ley]

I will be out of the office on June 8th.  If you need immediate assistance please call the main office number 503/223-8248 or 800/788-1887 and the receptionist will ensure that I get the message.  Thank you. Kelly