I conducted a survey in which surrounded three major themes: consumer behaviour, loyalty and satisfaction.
To analyse my data I placed the survey's data into SPSS to start with. I ran a Cronbach's Alpha on each construct, as normal I guess. Afterwards I wanted to examine if there was a correlation between the constructs, so I computed a variable for each one by calculating a mean score of responses for the questions which entailed that specific construct. This was then followed by running the Spearman's correlation test. It's safe to say the results have created a little doubt and confusion in my mind.
To give a brief overview, my 3 Cronbach's Alpha tests scores varied from 0.5-0.6 which in this rare case is acceptable due to the nature of this research being exploratory. Now the confusion arises as my Spearman correlations for one test sits at 0.88 at a significance of 0.000, meanwhile the other two sit in the 0.5 range. Does this seam feasible given the short, but brief information I have given here?
All help or advice would be greatly appreciated!
Dear David,
Before running a reliability analysis (using coefficient alpha) you should have performed an exploratory factor analysis plus a confirmatory one (using Lisrel or IBM's Amos).
Once you have confirmed the unidimentionality of the scales you can examine the correlations.
Hello Salim Moussa ,
I have performed the exploratory factor analysis as well as confirmatory analysis. Apologies I should have mentioned that! My issue really is when looking to run Spearman's coefficient, I created a new variable named Satisfaction for example into my dataset, based on the questions I used for the satisfaction construct. I have now confused myself however, as when I computed the variable, I am unsure if I should have computed a mean score of the variable to run, or should I have accumulated the total sum of each participant?
Many thanks, David
In response to your first question, it is quite feasible to have the pattern of correlations observed by you irrespective of the Cronbach alphas for the three scales used by you. This is because the reliabilities of individual scales has nothing to do with their correlations with each other.
As to your second question, once again it should not make a difference to the correlations among the variables if you calculate their average scores or sum totals. You may try this out to confirm.
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion