Hi Cyrielle,

How are you doing ? Fine, I hope.

Analysing your question, my uderstanding is that you could use Exploratory Factor Analysis + Multiple Regression or Confirmatory Factor Analysis + Structural Equation Modeling.

To suggest these two possibilities I'm assuming your 25 questions intend to represent some latent variables such as Music Experience and Social Experience, for instance.

So, if you want to study some cause - effect relationship, these two proposal should be quite useful.

Hope this can help.

If you need anything else, feel free to call me back.

Regards,

Thank you Alcides. I had in mind to proceed to an exploratory factor analysis + regression analysis indeed. To assess my first hypothesis, I've designed the following question for instance:

-"When attending a concert at music festival, a good quality sound equipment is important for you to feel a better connection with the artists"

1) Strongly agree 2)Agree 3) Neither agree or disagree 4)DIsagree 5) Strongly disagree

I did the same for each of my hypothesis. However I saw that Regression analysis was only suitable with interval/ratio data... Could it be possible to use it with a Likert scale?

Thanks for your help and time!

Kindly

Good morning Alcides

I had another question regarding the method to adopt to do my Exploratory Factor Analysis: Should I compare only variables in each facet or compare my 8 hypothesis at the same time?

Thanks

Hi Cyrielle,

Did you use just one question for each hyphotesis ?

If so, doesn't make sense to use EFA.

Let me use an example from Marketing area to ilustrate the whole picture.

Let's assume I want to know if Customer Satisfaction affects Customer Loyalty to a brand or a product.

So my hypothesis is that Customer Satisfaction affects positive and significantly Customer Loyalty.

Both variables are latent variables as both of them can't be measured directly.

To measure them I use two scales, one for each latent variable.

These escales use Likert range similar to the one you mentioned for your research (usually the format is: 1) Strongly disagree 2) Disagree 3) Neither agree or disagree 4) Agree 5) Strongly agree, but what you propose is not wrong).

Each scale has to have at least three statements

(Hair, Anderson, Tatham, & Black, 2009); for instance, for Customer Loyalty you could use: 1) I prefer to use the products of this company , 2) I think this company has the best offers in the present , 3) I prefer to buy this brand instead of other brands.

After that you should check at least internal consistency of each scale (Cronbach Alfa), but this is the minimum since in reality you should check also factorial validity, convergent validity and discriminant validity of each scales, which can be done when you make a CFA.

Assuming that the scales for both latent variables are OK, them you should calculate scores for Customer Satisfation and Customer Loyalty and them submit these data to regression analysis, checking if the hypothesis can be confirmed.

Despite EFA + regression analysis can be used, depending upon the magazine you want to submit it may not accept and ask you to perform CFA + SEM (I already have this bad experience with the magazine refusing my paper because ot methodology I used).

So, use one Likert-type question to test the hypothesis sounds a little weird, at least consideting what I have seeing in the literature.

Attached are some articles to ilustrate my example of Customer Satisfaction and Customer Loyalty (as I'm more familiar with this kind of research I decided to use it to show you the whole process).

Sorry if I were so basic in my discussion, but I'd rather to present the whole picture to permit you to confront my example with your situation.

I hope this can help you.

Anything else, feel free to call me back.

Bibliography for my discussion (the book has an English version):

Hair, J. F., Anderson, R., Tatham, R., & Black, W. (2009). Análise Multivariada de Dados (6th ed.). Porto Alegre: Bookman.

Dear Alcides,

Thank you very much for your precious help! I have proceed to Cronbach's alpha analysis and the results presented internal consistency for each scale.

You are then suggesting that I use Regression Analysis. I have read in literature that I indeed can inferred ordinal variables as interval variables. However I do not understand what you mean when you suggest that I should "calculate scores" before proceeding to the regression analysis?

Kindly,

Dear colleague, you may have a look to this outline. I find it very useful. The document is prepared by Dr Md. Mahmudul Alam and Mr Md Mahadi Hasan from Universiti Utara Malaysia.