Dear everyone,
I am a PLS Junior User and try to analysis a cross-scenario moderating effect. In order to assess measurement invariance, the MICOM was conducted through permutation function.
However, the results show that in step 2, original mean difference of one variable (0.9875) didn't fall within the 95% CI (0.9894), and in step 3, the Mean - original difference of another variable (0.3515) didn't fall within 95% CI (-0.2812, 0.2897)
The questions are:
1. The path coefficient of problematic construct is significant in both two groups. Can I remove it and analyze the reduced model in order to run the PLS-MGA? Following the instructions in the link viewtopic.php?f=5&t=15981&p=27553&hilit=MICOM#p27553, the step 2 is established, which means the PLS-MGA can be conducted.
2. In the remaining construct PLS-MGA results, all the path coefficients differences are not significant. Following this I can deem that there is no moderating effect of the different scenarios. However, in Boostrapping results, the p value of three path coefficients in Group 1 is lower than 0.05, while in Group 2, only one path coefficients' p value is significant. It's quite strange. How can I interpret this results?
3. According to Henseler et al, 2016 - " lack of compositional invariance implies that the scores obtained through group-specific model estimations differ from the scores resulting from the pooled data (no measurement invariance established). Researchers should only analyze and interpret the group-specific model estimations separately." If several paths' p value is significant in group 1 but not in group 2, can I directly assume that the significant difference is existing between these two group? That means the moderating effect exists between different scenarios?
4. For the step 3 results, how can I deal with or explain the problematic parts? I tried to find some references, but I failed.
Any clarification on these? any other resource please?
Thanks in anticipation.
Huan