Why is not possible to include both predictor, moderator and interaction effect in model?

53 views
Skip to first unread message

walkywalky

unread,
Aug 24, 2019, 10:23:03 AM8/24/19
to lavaan
Dear all,

I want to test the moderating effect of car use on the relation between effort expectancy and behavioural intention. 

I created the product term and now included this in my measurement model along with the predictor and moderator. Variables were mean centred before using the IndProd function.

However, I dont get all values for my model.

EE -IV
INT - Moderator

Regressions:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  BI ~                                                                  
    PE                0.276    0.078    3.532    0.000    0.337    0.337
    SI                0.102    0.052    1.972    0.049    0.106    0.106
    FC                0.329    0.144    2.275    0.023    0.304    0.304
    HM                0.388    0.100    3.899    0.000    0.398    0.398
    INT              -0.092       NA                     -0.388   -0.388
    EE               -0.311    0.131   -2.384    0.017   -0.295   -0.295
    CarUse            0.054       NA                      0.078    0.227

When I only include the IV and interaction effect, everything works!

Is it necessary to include both the IV, moderator and interaction effect?

Any ideas?

Thank you


Edward Rigdon

unread,
Aug 24, 2019, 11:35:41 AM8/24/19
to lav...@googlegroups.com
Very likely, you also got a warning message with your lavaan output, explaining that not allstandard errors could be generated. THat message might have given a reaon for the failure to generate standard errors. It is interesting that your output reports a p-value but not a standard error, since the standard error is an input to a p-value. 

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/b41da44c-bf21-402a-b005-72a5137d2421%40googlegroups.com.

Sina Nordhoff

unread,
Aug 24, 2019, 12:10:49 PM8/24/19
to lav...@googlegroups.com
Thanks for your help!

No, there is no error message.
Any idea what else can be done now?

By the way, when I use the output of the IndProd function and include the interaction term as part of my measurement model, the model does not run. 
In this model, I define INT as V2_EE1.CarUse + V2_EE2.CarUse + V2_EE3.CarUse

When I mean center the variables using this function scale(Data$v, center = TRUE, scale = FALSE) and then create a product manually (EE1 times caruse, EE2 times car use, EE3 times car use), the model runs except for the fact that it does not give the p-value.
Here INT is defined as:
INT =~ V2_EE1.c * CarUse.c + V2_EE2.c * CarUse.c  + V2_EE3.c * CarUse.c    




Terrence Jorgensen

unread,
Aug 25, 2019, 8:17:59 AM8/25/19
to lavaan
INT - Moderator

You mean this is the interaction term?  CarUse is the moderator, according to your other post.


Is it necessary to include both the IV, moderator and interaction effect?

Yes, otherwise you are not modeling moderation.



It is interesting that your output reports a p-value but not a standard error

That never happened.  You are seeing standardized-solution columns, which is hard to tell without setting the output's font to fixed-width like in the R Console.

Here INT is defined as:
INT =~ V2_EE1.c * CarUse.c + V2_EE2.c * CarUse.c  + V2_EE3.c * CarUse.c    

That lavaan syntax does not create product terms.  It just puts labels on the 3 factor loadings (e.g., INT=~CarUse.c is labeled "V2_EE1.c").

Unless you changed the default arguments of indProd(), the variables are already mean centered, and the product terms are double-mean-centered.  You can read about the arguments on the ?indProd help page, where you will also find 3 articles that explain how to actually fit the model correctly.  Read those before trying anything else.

Terrence D. Jorgensen
Assistant Professor, Methods and Statistics
Research Institute for Child Development and Education, the University of Amsterdam

Sina Nordhoff

unread,
Aug 28, 2019, 4:18:35 PM8/28/19
to lav...@googlegroups.com
Hey Terrence, 

Thanks a lot!!!! Really appreciated!
--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/253d4514-29f3-466d-980b-7209614fb5d7%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages