1
$\begingroup$

Can there be 2 types of measure in one questionnaire?

For example, both a 5-point Likert scale and a 7-point Semantic Differential are used in a single questionnaire.

If I were to compare the mean, should it be interpreted according to individual scale?

Background: In my questionnaire, respondents need to evaluate Website A and Website B whereby the constructs were assessed as follows: Question 1 - 24 were 7-point Semantic Differential Scale Question 25 - 48 were 5-point Likert Scale In my analysis, I applied Independent Sample T-test. I want to compare the mean values between respondents who evaluated Website A and respondents who evaluated website B. Being a newbie, I found that might be because I am combining 2 types of measurement scale in a single questionnaire. Is such a practice correct?

$\endgroup$
  • 3
    $\begingroup$ I would consider making the title of your question a little bit more descriptive. $\endgroup$ – Patrick Coulombe Dec 7 '15 at 17:11
  • $\begingroup$ Can you be more specific? I'm having trouble imagining a case where comparing the mean of a question that would be asked on the Likert scale vs one that would be asked on a semantic differential makes sense. $\endgroup$ – Jordan Bentley Dec 7 '15 at 20:38
  • $\begingroup$ In my questionnaire, respondents need to evaluate Website A and Website B whereby the constructs were assessed as follows: Question 1 - 24 were 7-point Semantic Differential Scale Question 25 - 48 were 5-point Likert Scale In my analysis, I applied Independent Sample T-test. I want to compare the mean values between respondents who evaluated Website A and respondents who evaluated website B. Being a newbie, I found that might be because Im combine 2 types of measurement scale in a single questionnaire. Is such way correct? $\endgroup$ – zaihasriah Dec 8 '15 at 3:29
  • $\begingroup$ Has everybody answered all 48 questions or are there 24 per website? $\endgroup$ – David Ernst Dec 18 '16 at 22:00
0
$\begingroup$

Yes, there can be two different types of questions asked in one questionnaire.

Of course, you can easily compare the mean of semantic differential question X for website A versus the mean of semantic differential question X for website B. You can do the same for a Likert scale, but that puts more stress on the interval nature of a Likert scale, which actually is ordinal.

But what you seem to want to do here is to treat the 48 questions as your observations and compare them with a single paired t test with n=48 pairs. That seems problemmatic not only because of the different scales, but because the scores on these questions will have variable amounts of correlation.

The way I'd approach this problem is to use a Principal Components Analysis on the raw data to determine how the questions are related to each other. Standardize the data first, to control for the different number of scale points. Then compute the component scores for the strong factors -- maybe just the first one if you want the single best measure. If you used the first principal component, you have one mean for A, one for B, and you use the variability of the scores on the raw data as your estimate of the variance. [similarly for the second component, etc. I can't tell without analysis how many components would be appropriate for this data. Check the eigenvalues.]

There are other ways to attack this problem, so you may attract other answers.

$\endgroup$

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.