I prepared moderated mediation model (education-intention-behavior relationship moderated by environmental factors including some demographic factors as control) for my upcoming research. I used popular and well verified likert scale variables. I just completed my pilot study with 45 respondents. The cronbach's alpha of all variables and constructs used in my pilot study generated good values (above 0.8). But when doing further analysis of pilot study, results reject most of the direct relationships: education-intention, education-behavior, intention- behavior, environment-intention, environment-behavior.
Thus, I think no need to do research on this issue because most of the hypotheses are rejected on pilot study. But, I'm confident on the model and scale based on literature review, the model and scales are well verified globally. As pilot study questionnaire was in English language, I think I need to translate questionnaire in local language, and done so for final study. At this moment, I'm confused to proceed the research.
Should I proceed this research? If again the results of upcoming final study rejected my most of the hypotheses, how can I address that problem in that time? What alternative ways are their to explain the results for rejected hypothesis? Or should I stop doing research in this particular issue. Or how can improve this research and questionnaire from this stage?
1) Most of variables in your study have not two values but they are multi-value variables. So, fuzzy logic will help you.
2) Not only education affects intention and behaviour, but also behaviour and environment have educational effects. So, a more complex casual network is needed.
3) Demographic should not considered as environmental variables. Re-conceptualize your variables through a new casual pathway.
4) Most of foreign language questionnaires have epistemic errors. Modify the questionnaire and publish such incompatibility.
More notes will available if I have more information.
A rejected theory or hypothesis dosnt imply that it is really wrong, that depends on how to explain results!. Regards
Don't forget that the small N in a preliminary study will affect your significance levels. The g*power program might help you assess whether a larger sample size would make a difference.
Don't forget that the small N in a preliminary study will affect your significance levels. The g*power program might help you assess whether a larger sample size would make a difference.
Don't forget that the small N in a preliminary study will affect your significance levels. The g*power program might help you assess whether a larger sample size would make a difference.
Insignificant results might point to mediating or moderating effects which you have missed. Thus, there might be intervening variables which decide whether a significant relationship exists or not. Generally, and in line David L Morgan, the smaller the sample the more difficult it will be to find significant relationships. So a larger sample might actually yield significant results.
- Badges
- Science topic
- Similar topics
- Psychology