Factor analysis is most useful for assessing scales, i.e., multiple items that all assess related concepts. A scale can be an independent, mediating or dependent variable, but you cannot combine those different functions into a single scale.

If you have ideas about variables being independent, mediating, and dependent, why do you want to do an EFA? Have you drawn out a causal diagram (a graph) of your beliefs about the causal relationships among your variables?

Factor analysis is most useful for assessing scales, i.e., multiple items that all assess related concepts. A scale can be an independent, mediating or dependent variable, but you cannot combine those different functions into a single scale.

Mazhar,  in factor analysis, the issue of dependent and independent variables doesn't arise.  By its  name, factor analysis classifies or brings together all the highly correlated variables suppied  into the same  factor. Many  factors may emerge depending on their correlations. Each of the factors usually give a clue to the name   for  the factors.

The basic purpose of EFA is to develop a questionnaire and identify true dimensions in a concept. The EFA may guide you to identify IV, DV and MV but if you already know your variables and you have adopted questionnaires then use CFA for contextual validation of your questionnaires and for that purpose use AMOS, draw a diagram and get the result. If you already know that these variables have correlation then you may face problems of multicollinearity. 

Dear Mubashar thanks for your contribution. Some  researchers often arbitrarily classify the various factors or aspects of an issue  by their intuition. However,  a high  level research statistics like factor analysis, needs to be  used to confirm the human opinion displayed by the researcher.  Factor Analysis uses only  the correct  correlations between  all the variables in the analysis.

If the number of variables is large for both IV and DV it is better to run factor analysis separately. There is another problem of including all variables at once i.e If IVs is an ordinal scale and DV 's are dichotomous further it complicates for normality, variance extracted and model fitting.

My understanding of exploratory research is that not much is known about the variable you are studying and so you want to generate basic information about it. If, however, you are able to generate dependent independent, and moderating variables, then you can do factor analyses of all these variables to increase their understanding without necessarily measuring any explanatory or predictive power. To attempt assessing explanatory or predictive power be going beyond exploration into relational study which may lead to you spreading out too thin for effectiveness,

Hello Mazhar,

I think that the first two comments above, made by Daniel Wright and David Morgan, address the main issue you are confronting - namely whether or not performing factor analysis by combining your independent, dependent, and mediating variables would be a useful exercise.  If I read Daniel and David's contributions correctly, they are indicating that it would be misguided to do so - and if that is what they are indicating, I agree with them.  Factor analysis can be extremely useful, but it should be used only in appropriate contexts - and combining all of your variables into a factor analysis would not be one of those contexts.  I suspect that it would confuse, rather than clarify, the issues you are attempting to address.

It might be advisable to factor analyze your independent, dependent, and mediating variables separately, but you have not provided sufficient information for others to advise you concerning whether or not that might be a valid or useful tack to take.

All the best with your research:  Go forth with clarity!

Robert.

hi, my question is, can we use anova or t test on using IV, DV and mediating variables??? we have run mediation analysis...now we are looking for analysis on secondary hypotheses. so my question is, is it theoretically acceptable to use anova or t test by taking IV, DV and mediating variables simultaneously???

Dear Kazmi, T test is actually conducted;

• when your research question is "Is there a significant difference in the mean DV scores for males and females". It will tell you whether there is a statistically significant difference in the scores of two groups (males and females).
• To conduct T test you have to have one categorical independent variable (males- females). In case of continuous variable you need to split such into two categories.
• one continuous dependent variable.

Dear Ali.

To my understanding. Its depend on the nature of your variables with their underlying structure.

For example if you have service quality (IV) with five dimensions, you need to check the underlying structure in different setting for this variable separately. In case of first order factor you may run EFA at the same time for all variables.

Note. Nature of underlying structure of items need to be considered while conducting EFA.

Hello everyone,

Do you any journal stating that dependent variable can be separately from independent variables by using EFA?

Tithtola, there is no need for a journal citation on this issue. The independent variables always have to be kept separate from the dependent variables, so that one can predict the other. if you have multiple measures for both your independent and dependent variables, then you need to do two factor analyses, one for the measures of the independent variable and one for the dependent variable.

Basically, there is no need to separate such a variable in EFA: Article Korean Patient-Perceived Satisfaction Scale of Community-Bas...

If you would like to develop an instrument and test its psychometric properties, you need to separately perform the factor analysis; that is, one for the independent variable and one for the dependent variable.

David L Morgan thanks for your comment. No doubt a novice question but I myself was wondering if you run separate analyses on the IV and DV items. As the OP said in his original post, intuitively you would want to separate them.

Tithtola トラ Nop

Hair et al. (2010, p. 99) stated:

" For example, mixing dependent and independent variables in a single factor analysis and then using the derived factors to support dependence relationships is inappropriate."

Hair (2015, p. 411) stated:

" When you use exploratory factor analysis the variables are not divided into dependent and independent categories. Instead, all variables are analyzed together to identify underlying patterns or factors. The technique can be used to factor analyze either independent or dependent variables considered separately "

Hair, J. F., Black, B., Babin, B., & Anderson, R. E. (2010). Multivariate Data Analysis 7th Pearson Prentice Hall. Upper Saddle River, NJ.

Hair Jr, J. F., Wolfinbarger, M., Money, A. H., Samouel, P., & Page, M. J. (2015). Essentials of business research methods. Routledge.