HI Anna,

A moderator variable is nothing but another predictor variable.

In your case, you run a multiple regression with your three subscales (egoistic, altruistic and biospheric, level: metric) + your fourth subscale (moderator) as predictors and metric as your dependent variable.

You simply compute (observed) sum scores for each predictor by adding indicator variables' observed scores. In this case sum scores of you latent variables (egoistic, altruistic and biospheric, level: metric, moderator) can be easily estimated by summing observed score of the relevant indicator variables (that are connected to "egoistic", "altruistic",... in your SEM). Hope this is sensible to you.

Cheers, S-K

Thank you very much for your reply. I'm currently working with PROCESS. Please correct me, but it only allows analysis for one independent variable, no?

That chapter seems really helpful, thank you.

If I have more than one predictor and one of them moderates the effect between the predictors and the outcome, as I pressume, than I need to perform a moderator analysis within multiple regression. Or am I wrong?

Anna,

you already got some helpful suggestions how to proceed. Just a few thoughts:

If you would like to estimate an interaction effect using the moderated regression analysis, you would have to chose which subscale you would think would be most important for your model. You would have to select for each construct only one scale, resulting in two predictor variables (one is the moderator variable) and one dependent variable. Then you have to form the product of predictor*moderator and enter this variable as the third predictor into the regression equation.

There is a nice book which I like very much and in which you will find all steps to perform a moderated regression using SPSS:

Field, A. (2013). Discovering Statistics using SPSS (4th ed.). Thousand Oaks, CA: Sage.

Older versions of this book would be fine, too.

Of course it would be better to analyze a moderated structural equation model, because only then you would be able to make consistent use of all information available.  If you are familiar with one of the SEM programs, just tell us, I guess somebody would be able to help you with the details.

Kind regards,

Karin

HI Anna,

A moderator variable is nothing but another predictor variable.

In your case, you run a multiple regression with your three subscales (egoistic, altruistic and biospheric, level: metric) + your fourth subscale (moderator) as predictors and metric as your dependent variable.

You simply compute (observed) sum scores for each predictor by adding indicator variables' observed scores. In this case sum scores of you latent variables (egoistic, altruistic and biospheric, level: metric, moderator) can be easily estimated by summing observed score of the relevant indicator variables (that are connected to "egoistic", "altruistic",... in your SEM). Hope this is sensible to you.

Cheers, S-K

Hey Anna,

I have the same problem like you. I have a multiple Regression with 4 independent variabls and one of the independent variabels is moderated. Actually i cant really see an conclusion about the answers. Because they just looking at a simple regession and made through the moderater to an multiple regression.

Did yu had a conclusion? I was just thinking turning the multiple regression in 4 simple regressions and then i can do the moderation effect. But not sure if then iam missing the effect of all together.

best

Milena

support Karin you can get the answer from the book of Field

Hi.

I am new to research. I have 2 independent variables (Push and pull factors), a mediating factor (destination image) and a dependent variable (destination choice). However i have a moderator (perception of partner's preference) which has another data set. Can someone help me in how to run my analysis with the moderator.

Reagrds