how to carry out moderation in SPSS?

#1
I am carrying out a multiple regression on well-being but as well I would love to check

if Sense of Control (moderator) moderates the relationship between Social Status (IV) and Well-being (DV). Could you help me please? All of these are interval. I have been looking online but it is really hard to find something, I am quite surprised.

Thank you
 
Last edited:

Lazar

Phineas Packard
#2
I do realize that this topic might have been discussed already but so far I could not find anything. I am carrying out multiple regression on well-being:

I would love to check if Sense of Control (moderator) moderates the relationship between Social Status (IV) and Well-being (DV). Could you help please?

Thank you
COMPUTE contXSS = Sense of Control * Social Status.
EXECUTE.

use this as a predictor in your model along with sense of control and Social Status.
 

spunky

Doesn't actually exist
#3
and dont forget to centre your predictors to solve the multicollinearity of the interaction term and do the simple slopes analysis if the interaction (moderation) is found to be significant! :p
 
#4
COMPUTE contXSS = Sense of Control * Social Status.
EXECUTE.

use this as a predictor in your model along with sense of control and Social Status.
Thank you, I understand that you are an expert but I am not sure what contXSS means.

COMPUTE VARIABLES... an then I am lost.
 

Lazar

Phineas Packard
#5
Thank you, I understand that you are an expert but I am not sure what contXSS means.

COMPUTE VARIABLES... an then I am lost.
contXSS is just a random name I made up to call your interaction variable (cont = sense of control, x = multiplied by; SS = social status put it all together and you get contXss). You can call it what ever you like so long as it represents the product of your sense of control variable and your social status variable (as spunky says don't forget to center your variables first).


You can also just use the drop down menu. Go to transform > compute variable. Give your new variable a name (whatever you want) then in the big box put your sense of control variable then the * symbol for multiplication then the social status variable.
 
#6
Thank you for that.

But actually I wanted to check if:

Sense of Control will moderate the relationship between Social Status and Well-being. This is a second hypothesis that I have and I would not like to include it in the main model. Any suggestions how to check it?
 
Last edited:

spunky

Doesn't actually exist
#7
well... you kinda do what lazar already told you to do.
since you only say "relationship betweens status and well-being" let's assume (wlog) that your predictor variable is status and your criterion variable is well being... so, steps:

1) create new variables centering sense of control and social status (i.e. you substract the mean of each from every single data point in your data-column)
2) create the interaction effect (from the centered data) just as lazar said, multiplying both your predictor (social status) and your moderator (control)
3) run an ols multiple regression with and without the interaction term. if the interaction term is statistically significnat, then you can conclude that control is moderating the relationship between status and well-being (and i know a few people like to check whether the increase in R^2 is also statistically significant or not, which would also bring more evidence in favour of the moderator effect).
4) if the interaction is not significant then you can interpret the main effects only. if it is significant then you should run a simple slopes analysis to understand the interaction...

this is an extremely brief summary of what is usually an, alebeit not complicated, kinda cumbersome process (at least to me. i dont really like moderated regression). i'd really suggest for you to take a look at the classic of classics Cohen, Cohen, West & Aiken's book on regression/correlation analysis.