I am doing undegrad thesis sand I stumble on some major problems when doing factors analysis:

My research topic is to compare the attitude of people towards 2 types of products (namely A and B).

My naive original plan was to create then compare 2 homogenous SEM models for each type of products and the proposed model is based on prior researches as well as my own rationale, so I guess I would have to perform EFA on one part of the data and the CFA and SEM for the rest (N=150).

However, when doing EFA, the variables loading in unwanted way and for each type of product, each set of data has a different pattern, thus, it is impossible to carry out my plan.

Now, I am thinking of merging the answers for 2 types of products into 1, since I have 2 set of similar questions asking the same aspects but of different objects (A or B). Additionally, there would be another column to distinguish 1-for product A and 0-for product B.

Pro: With an unified poll of data, I will only need to make one model.

Cons: I am not sure yet how to make the comparison if doing so.

I cannot tell which 2 opinions come from whom (which I don't think necessary).

And, instead of having N =150, it would be N=300, I don't know whether this will lead to false statistical results.


1. CAN I SKIP DOING EFA and jump on CFA right away?

2. If not, will my solution of merging fix my problem?

3. If not gain, what else can I do?

Thank you in advance!!!


1 Answer 1


Merging the data (N=300) looks a better way. So you can build a structural model to regress attitude on product type (a dummy variable, or a single-item construct). This allows you to easily explain the difference in attitude towards different product types.

For your questions: if the measurement items have been widely used before, doing CFA should be fine. If the items are self-developed, you have to do EFA.

  • $\begingroup$ thank you for your answer. I have had a second thought on the merging solution, now that I find it problematic since when I merge 2 columns of data into 1, the combine column would not carry any particular meaning. Like if one person think the quality of product A is 5 and B is 1 then we cannot say the quality of A or B is 3... so I think I have no choice but to skip doing EFA?! $\endgroup$
    – KBV
    Apr 28, 2021 at 8:13
  • $\begingroup$ So you used a within subject design. This makes the SEM more complicated. Even the sequence of products A and B might matter. A simple way is to do CFA or EFA for each product, then calculate the mean of all attitude items for each product after CFA/EFA. Then conduct an ANOVA with repeated measures to test the difference in attitude towards A and B. $\endgroup$
    – Alan
    Apr 28, 2021 at 17:31

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.