Analyzing the moderating effects for the model with latent constructs is very complicated. The normal modeling procedure using interaction terms is not practical with latent constructs since it would cause problems with model convergence as well as distortion of standard errors. In the end, it resulted in model misfit and the procedure stops.

Figure 7 illustrates how the moderator is modeled when analyzing the model consisting latent constructs.

Figure 7: Modeling the moderator variable in the path between X1 and Y

Alternatively, the Multi-Group CFA has been suggested as an alternative method for assessing the effect of moderator variable in the model. The researcher only needs to identify the path of interest where the moderator variable is to be assessed. This particular path would be constrained with parameter = 1 and the model is termed as the constrained model. The procedure will estimate two models separately. One is the constrained model while the other one is the unconstrained model. The step by step process for Multi-Group CFA is discussed.

Yes, you can perform multi-group in other softwares too like SPSS, R, Mplus, and so on. I'm not sure what exact analyzis are you trying to perform, but why not starting with some tutorials? There are some good ones in youtube on multi-group in AMOS, for example:

https://www.youtube.com/watch?v=H4gHIs4y6sU

Good luck!

Analyzing the moderating effects for the model with latent constructs is very complicated. The normal modeling procedure using interaction terms is not practical with latent constructs since it would cause problems with model convergence as well as distortion of standard errors. In the end, it resulted in model misfit and the procedure stops.

Figure 7 illustrates how the moderator is modeled when analyzing the model consisting latent constructs.

Figure 7: Modeling the moderator variable in the path between X1 and Y

Alternatively, the Multi-Group CFA has been suggested as an alternative method for assessing the effect of moderator variable in the model. The researcher only needs to identify the path of interest where the moderator variable is to be assessed. This particular path would be constrained with parameter = 1 and the model is termed as the constrained model. The procedure will estimate two models separately. One is the constrained model while the other one is the unconstrained model. The step by step process for Multi-Group CFA is discussed.

You can indeed also perform moderation analyses with SPSS, or SAS for instance, Depending on which kind of model you would like to test, you can decide for doing an SEM with e.g. AMOS or go for the regression based analysis with SPSS. Generally, SEM is a more sophisticated method if you have more than one dependent variable or mediators as it tests all hypothesised relationships simultaneously. An additional advantage is the fact that it includes measurement errors, which regression- based analyses in SPSS don't. 

One point of attention when doing multi-group analysis in SEM is to ensure measurement invariance between the two groups before you do your analysis. 

Good luck!

Read this attached chapter (Chapter 7) from my book titled A Handbook in SEM for Academicians and Practitioners.

Dear Zainudin and Stephanie,

Thank you very much for your meaningful and valuable insight on this issues.

Now I am bit confidence on my hypothesis testing.

Once again thank you very much....and share your expertise as much as you can.....

There is also this book by Barbara Byrne, its name is Structural Equation Modeling with AMOS", it also shows many applications available with amos software

I want to do multigroup moderation. how can I do it in SPSS

Fazli, SPSS (Regression) is not meant for latent constructs. It cannot analyze the moderating effects between latent constructs. In traditional SPSS, you have to convert scores from measuring items into mean score and analyze it as observed variables (no longer latent constructs). In fact this is one of the limitation in SPSS-Regression where SPSS-AMOS can cater. In SPSS, we are not analyzing moderation effects but interaction effects. Moderation effects in latent constructs is interaction effect in observed variables. You need to model the interaction between IV and Moderator and test the hypothesis for its beta coefficient.

If i have six group of industries and If i make 5 dummies and multiply with the main independent variable. would it be wrong ? what is the difference between moderation and interaction.

It is the same thingdepend on the types of modeling you are doing. If you are modeling the observed variables using regression, then you call it interaction effect between independent variable and moderator variable.

If you are modeling SEM since you are dealing with latent constructs, then you call it the moderation effect of the moderator variable in the relationship between Exogenous Construct and Endogenous Construct.

You can see http://www.amazon.es/Structural-Equation-Modeling-With-AMOS/dp/0805863737

Hi:

I am going to collect data from 5 construction sites to check the relationship and impact between/of different factors (safety climate, psychological contract, safety behavior and safety outcomes). Three of these four factors will be measured by perceived Likert Scale and another factor safety outcomes will be measured by a numeric scale. The numeric scale will be developed through a weighted scale (please see attached) from 1-10. All the other perceived scale will be used also from 1-10 likert scale. But in case of numeric scale I will have only 5 set of data (safety outcomes scale data from 5 construction sites) whereas for other factors I may have 400-500 data set from workers on 5 construction sites.

1. Do you think I can do acceptable analysis and result from these data sets?

2. What are the analysis I can do and could you please give me some references where this type of study is conducted?

3. I will have around 50 questions in my questionnaire for the first three factors. Is it alright to have this number of questions for SEM analysis?

Thank you so much for your patience reading. Looking forward to your answers.

Tanvi

You can use AMOS, LISREL or Mplus to conduct multigroup analysis, for amos there is this paper from barbara byrne on Factorial invariance that might be very helpful:

Byrne, B. M. (2004). Testing for multigroup invariance using AMOS graphics: A road less traveled. Structural Equation Modeling, 11(2), 272-300.

Invariance testing requieres multigroup analysis with equality constrains (in case that you want to test invariance in CFA). If want only to perform multigroup only is way easier, only consider that you need at lest 300 cases for each group, if the model is relatively simple, however if its comple you will need >500

@Mohammad Tanvi Newaz: 

First, is would be highly recommendable that you post your questions independently, so you might have more chances of a good answer. 

1. The answer for this depends more on the theoretical rationale underlying the analysis.

2. It seems that you are conducting a SEM model, try to use estimators from the WLS family.

3. Absolutely not, a questionnarie using 50 indicators is too much for CFA (a special type of SEM, which is nested in the whole SEM model), CFA is too much restrictive so is highly possible that your model fails to be identified.

Hope this will help you Mohammad!

I have a group with 91 and a group with 208.....can I do anything or my sample size is not enough?

a group with 91 subjects is just too little, remember that assumptions in CFA are very restrictive, in my experience to assess invariance using multigroup analysis in SEM you need at least 300 by group, however you must take into account the number of items, the measurement error, the multivariate normality (which is very, very rare) and the amount of missing data.

 I am using AMOS to analyze multigroup, but I don't understand what does it mean, the model 1, 2, 3,..., 8 in the logiciel. Could you give me some documents or instructions?