Dear all,
I hope you can give me your opinion about this.
I have to test the following relationships (see attached image):
X1 to X4 are my main independent variables; they cannot be measured directly, but consist of the antecedents variables e1 to e8. Basically, I want to see the direct effect of X1-X2-X3-X4 (basically e1 to e8) on the dependent variable Y.
After that, I want to test if and how two moderators variables ( X5 and X6, both categorical, specifically male/female and individualist/collectivist) affect these relationships.
What I want to see, after the main direct relationships, is comparing these relationships between different groups (e.g. effect of X1 (e1-e2) on Y comparing the men sample vs women sample; individualist sample vs collectivist sample; men vs women only in the individualist sample; etc.).
I have some experience with multiple regression; however, looking for some information, I found that many analysis with similar conceptual model use SEM to test these relationships.
What is your opinion about it? What would be the best method?
Thank you in advance for any help.
If more recommended is the SEM analysis moderation although MIMIC SEM model can also determine differences by groups.
Giuseppe Alfano
I recommend you to use SEM analysis using SmartPLS3 for your desired estimations.
Zahid Bashir Thank you for the suggestion. During my university studies I learned a bit to use Stata; do you think it could be manageable to use it to perform SEM?
Giuseppe Alfano
Stata can be used for SEM, but I personally recommend SmartPLS3 for SEM as it is easy to use and it depicts results in a much better way then stata. I personally use STATA when I have to estimate secondary data.
Zahid Bashir I understand, I'll try to use SmartPLS3 then. Appreciate your help!
If you want to compare different groups ANOVA or two sample tests are more suitable. Furthermore you may use dummy variables to check which factor of group affects more. Its quite simple to run on SPSS. You can use it as additional analysis along with smartPLS.
Naveed Hasan Thank you for your opinion Naveed. I did Anova and two-sample test on Stata in the past. However, does it make sense to do that with latent variables such as in my case?
Unfortunately I never managed latent variables, that's why I'm asking.
It depends on your study objectives. Your complete model can be analysed through SmartPLS or AMOS. Additionally if you want to check the which factor of gender or Culture affects more (or which affects less), that could be done using ANOVA or two sample test.
Naveed Hasan The main hypothesis regard the direct effect between X1-X4 on Y; then, I want to test these direct effect comparing the different groups, to understand if e.g. the direct effect of X1 on Y is stronger for male or female.
The 'problem' is that X1-X4 are not directly measurable, but they are composed of the subcomponents. To make it easier I thought about calculating the average of the subcomponent to assess the value of X1-X4, but it would obviously not be a great analysis because they don't have all the same weight.
I recommend to use AMOS or smartPLS for direct relationships and SPSS or Stata software for comparison purpose.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude