2
$\begingroup$

I have designed a questionnaire which has 16 Likert item questions; from 1 (strongly disagree) to 5 (strongly agree). The 16 Likert item questions are divided into 2 groups: 10 questions for customer loyalty, and 6 for customer satisfaction.

  • How do I know whether the Likert items should be grouped into the two scales?
  • How can I calculate the relationship between the two scales?
$\endgroup$
2
  • 2
    $\begingroup$ There are hundreds of questions on this site about similar data. Did you look at any of them? Search [likert] and see stats.stackexchange.com/questions/31598/… for one example. $\endgroup$
    – Nick Cox
    Aug 20, 2013 at 9:09
  • $\begingroup$ @NickCox Actually most of the questions are about Likert-type items, this one is one of the few using the term “Likert scale” in its original psychometric sense. There aren't too many questions about that. $\endgroup$
    – Gala
    Sep 16, 2013 at 13:01

3 Answers 3

2
$\begingroup$

You probably want to do exploratory (or possibly confirmatory) factor analysis to check that all items can be scored in terms of your two theorised scales. This could be combined with reliability analysis. You would then score your two scales and then obtain the correlation between the two scales.

I have a tutorial which describes this process using SPSS including scoring the scales.

If you wanted to study latent variables and the measurement model of the test you could use structural equation modelling software. If you were in the world of SPSS, this would involve using Amos.

$\endgroup$
1
$\begingroup$

Factor analysis (which in SPSS will actually run a principal components analysis by default) is what you are after. Or, if you want to test that your theoretical grouping is appropriate you can use confirmatory factor analysis (historically SPSS have marketed this as a part of their AMOS module, but I am not sure if they still do this).

Most of the more advanced market research texts will contain worked examples dealing with how to do this.

$\endgroup$
2
  • $\begingroup$ I'd water this down a bit. These are very widely used methods here, but by no means the only ones. Just averaging ratings of one kind or another remains a simple and often used alternative. $\endgroup$
    – Nick Cox
    Aug 20, 2013 at 13:58
  • $\begingroup$ Simply averaging ratings is not an alternative to factor analysis. It just assumes the items can be treated as a scale. $\endgroup$
    – Gala
    Sep 16, 2013 at 13:05
0
$\begingroup$

Perform Reliability test such as Cronbach's alpha. Also perform Factor Analysis and inter item correlation with deleted items. Many softwares are available for perform these test, SPSS is easy to used.

$\endgroup$

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.