1
$\begingroup$

I am doing a master's thesis on the roles played by Perceived Bank Image, Perceived Service Quality and Perceived Satisfaction in predicting Customer Loyalty. My questionnaire consists of several questions for each of these constructs. A 5-point Likert scale is used to measure responses for each question.

I would like to conduct correlation and regression analyses to determine the strength of these relationships in SPSS. The process is pretty straightforward for constructs with a single question. With multiple questions however, I do not know how to properly proceed. Should I compute a composite variable, or should I simply analyze all the sub-items together?

$\endgroup$
  • 2
    $\begingroup$ Both strategies could be sensible and illuminating. $\endgroup$ – Nick Cox Nov 20 '15 at 18:16
2
$\begingroup$

If you have several questions for a construct, a structural equation model could be appropriate. You would need the AMOS module since SPSS advanced stats does not carry it.

For a cheap and easy workaround, summarize the constructs with a score. You could take the mean, or a factor score from the factor analysis menu.

If you have no idea what I am talking about, find a statistician or a psychologist to sit down and talk you through some basic measurement theory.

$\endgroup$
  • $\begingroup$ Thank you for your reply! Since my SPSS resources are rather limited, I will opt for summarizing the constructs with a score. Would the outcome be any different if I took the mean vs. factor score or vice versa? $\endgroup$ – Let's Do this Nov 21 '15 at 11:07
  • $\begingroup$ the mean gives the same weight to everything. the factor score downplays variables that are not strongly correlated with the others. Either way, you want to first standardize the variables by dividing by their standard deviations. $\endgroup$ – Placidia Nov 21 '15 at 14:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.