Can anybody help us regarding Exploratory factor analysis.
Is it possible of doing Factor analysis of moderating variable and dependent variable separately?
Abhiskek Shukla,
Your question is not very clear. EFA is used to reduce variables and get a first glimpse of a theoretical model.It is up to you to choose which variables you will use to do EFA.
If you wish to check the relationship between separate variables, maybe you should use analysis of variance?
FA makes no sense for MVs. You choose the moderating variables, not theory. By your choice, you are proposing no direct relationship with DV. Moderating effects should ideally be examined after you construct factors, ensure that they are as non-overlapping as possible, and explain a reasonable portion of variability in DV. Once the factors are identified and theoretic explanations are generated, then moderating effects on factors should be examined.
Exploratory Factor Analysis {EFA} deals with items in scales. The main function for EFA is to identify the intercorrelations among group of items call them by the end factors. Researchers sometimes, use scales such as Likert format, and get the total score from these scales, then use them as DV or Moderator. if this is the case your study, then you can analyse scales under EFA.
Dear sir,
Thank you for the response. Actually in my research model we have one moderating variable and some dependent variable in which having some dimensions. But after the survey we are getting good EFA, but when we are going to do CFA by AMOS, we are not getting the same results as EFA shows. So we thought might be we have to analyse EFA of all moderating and dependent variable together.
Can anybody help me out how to proceed further.
Abhishek Shukla
EFA is carried out to determine the usefulness of items in measuring the respective construct (one construct at a time). The usefulness of items is reflected through factor loading. Another purpose is to determine the dimensionality of items in measuring the respective construct. You have to do EFA for each construct separately.
I think you are still confuse between EFA and CFA. You see EFA is carried out using Pilot Study data whereas CFA is meant for Field Study data. You said your CFA results differ with what you get from EFA? Well since EFA and CFA were carried out for different purpose, using two different set of data collected at two different time intervals namely Pilot study and Field study, and using two different sets of algorithm namely OLS and MLE you should expect the difference where certain items need to be deleted and certain items need to be constrained. However, the deleted item in CFA should not exceed 20% of items after EFA.
Adding to Zainuddin, it is better to take a large single sample when planning to do both EFA and CFA. A smaller part of the sample (30%-40%) is selected using Monte Carlo methods, and EFA is performed on it. Then the remaining sample is subjected to CFA and the results are compared statistically. That way you do not have to justify sampling bias.
Dear Abhishek,
Your question is not clear. First of all, you should be clear why do you need to do EFA? It is not the nature of a variable (i.e., dependent variable or moderating variable) in a particular study that necessitates EFA. EFA is normally conducted for a variable to acertain its dimensionality if there is no theoretical basis for it.
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments