I have two Constructs (composite variables), each Construct has its own measurement scale (questions). First Construct is named CETSCALE and it consists of 17 questions, second Construct consists of 6 (or 5) questions (PERC_QUAL).

I want ot test if "CETSCALE" somehow influences "PERC_QUAL". How to test this? What tests to do? I would prefer to use SPSS, but for now I just want to know which tests to do.

  • 2
    $\begingroup$ I am puzzled by what you call an item (a question with multiple response options) or a scale (usually composed of several items with identical or different response options, and for which we can derive a total weighted or unweighted score). Could you clarify? $\endgroup$ – chl Jan 7 '15 at 15:43
  • $\begingroup$ I am very new to data analysis, so I can use incorrect names. I changed item to CONSTRUCT and it is scale that composes of composite several items. $\endgroup$ – renathy Jan 7 '15 at 15:56
  • $\begingroup$ I assume what you really mean are composite variables, correct? IE, your first, CETSCALE, is 17 questions, and then you take the mean of the scores to those 17 questions or some other calculation that arrives at a single score? $\endgroup$ – Joe Jan 7 '15 at 18:36
  • $\begingroup$ Yes, you are right. $\endgroup$ – renathy Jan 8 '15 at 9:15

This is pretty simple - just include a regression in your SEM model PERC_QUA on CETSCALE and use a T-test for that regression coefficient.

  • $\begingroup$ Can you please point me more in details or some reference how to do this in AMOS. I am new to this and never did SEM and used AMOS. $\endgroup$ – renathy Jan 8 '15 at 18:14
  • $\begingroup$ I don't use amos. In Mplus you can add the command PERC_QUA on CETSCALE $\endgroup$ – TA72 Jan 8 '15 at 18:20

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.