Hi there,
Currently, I am writing my thesis on how employees perceive different HR bundles (IV) and how this influences their level of engagement (DV). I am particularly interested in exploring the potential differences between different type of employees (MOD 1) and employees working for different type of firms (MOD 2). Please see the table below for a complete overview of my variables and their respective type of measurement.
I managed to collect survey data of +450 employees. Unfortunately, I do not have any solid background in conducting academic research, and I am struggling a lot with how to handle the statistical analysis part. (Except assessing the validity/reliability of my scales with Harman’s test; Cronbach’s alpha; VIF; AVE; CR; R2; fit indices etc.)
The past weeks I have reviewed a lot of literature on this topic, and I now believe I have the following options:
1. Treat my IV’s and DV as continuous (even though it is debated that Likert should be treated as ordinal) and conduct “multiple linear regression”,
2. Treat my IV’s as ordinal and DV as continuous/categorical and conduct “ordinal logistic regression”
3. Treat both my IV and DV as 3 factors measuring 1 latent variable and build a “SEM” model based on this, then conduct a multigroup analysis.
I want to research if firm type and employee origin are moderators, and therefor I aim to investigate if there are significant differences between the different groups within my study (locals in domestic firms, locals in multinationals, expats in domestic firms, expats in multinationals). Would you recommend me to research the moderation effect with one of the approaches described above, or with a completely different method? (I preferably make use of Stata, AMOS, R or SPSS)
I hope I have provided you with all important details, if not I am happy to provide more information. Thanks in advance for advising me, I am afraid you are my last hope of understanding which analysis I should perform and why. Regards Sophia
Since your DV is ordinal I suggest that you consider ordinal logistic regression. @David L Morgan has an excellent publication on creating ordinal scales which you may need..He is available on RG. An introduction to ordinal logistic regression is attached. Much more information can be had by Google search. Best wishes David Booth
BTW I forgot to mention that the material in the attached screenshot may also be useful. Best wishes David Booth
Hello Sophia,
in a SEM framework, you will not have problems with the 7 point rating scales, however, you will run into issues when forcing your variables into a common factor structure, as the variables unlikely correspond to such a structure but rather to an aggregate construct.
Perhaps applying PCA to reduce the overall number to correlated sub-HR functions but at the end you will probably have one or more composites and you cannot really say what the relationships with other variables will look like. In addition, your model lacks any control variables. Perhaps your survey also included these, too and you can add them to the model.
Good luck.
Holger
Hi,
Multi-group moderation analysis might help (SEM).
Best regards,
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS