Should correlation table findings inform what I test for moderation in a regression analysis?

Repi

New Member
#1
Hello there,

I am in bit of a trouble as I have never in my life done statistics and I need to analyse complex data for my research. I have SPSS and Youtube has been a massive help but I am unsure about whether correlation should inform the moderation effects I am looking for in a regression analysis.

My research is looking at the impact of negative acts in the workplace on employee engagement and retention. For this, I have prepared the survey which listed 5 behaviours (receiving unfair feedback, having misstatements made about you, etc.) About each of these behaviours, I asked 2 questions: How often do you experience these (which is ordinal data which I then coded like 1=never.....4=all the time) and to what extent do you believe these to amount to bullying (1=this is not bullying......4=this is bullying in any case).

This means that I have 10 variables (2 for each behaviour) which I then have to perform a regression analysis on for each question measuring engagement and retention... So far I have run regression analysis the following:

- All 10 independent variables against each dependent variables
- All 10 independent variables plus the 5 'products' for each behaviours between the respective experience and interpretation for all against each dependent variables to see the moderating effect
- All dependent variables separately and in pairs for experience and interpretation

However, I have done a correlation analysis as well and experiences for some of these have statistically significant correlation. If I run a regression analysis using two correlating experiences for two different behaviours and their product is it more likely that I will get a statistically significant result in the regression analysis? If I don't get a significant result how would I explain it? Would it mean that people who experience one tend to experience the other one as well but that these experiences have no impact on their intention to leave? Also if there is no correlation, there still may be moderating effect possibly so if I run it for strongly correlating ones, I should also run it with weaker ones or even ones that are not correlating significantly, right? Is there an easier way to find out what combination of experiences have the strongest positive moderating effect on engagement and retention without me having to run the test for every possible combination?

Thank you for your help in advance!