# How to run regression analysis without extracted factors from factor anlaysis?

I used Oblique Rotation in my Factor Analysis to reduce the dimensions and to extract 4 factors. But Since I was using Oblique rotation, the results of Factor Analysis did not contain the extracted factors' dataset that I can use in the following regression analysis.

How can I run the regression analysis without using the extracted factors' data? Should I calculate Z Score of each factor or question items to run regression analysis? CONFUSED.

Thanks.

I'm confused what you're confused about. If I understood your question correctly, your plan is to perform regression analysis, using factors, extracted during exploratory factor analysis (EFA). Let's assume that your original data set contains $N$ observations and $k$ columns, equal to the total number of factors. Your EFA resulted in 4 extracted factors (not the corresponding data, as you rightly noted), let's call them $f_1, f_2, f_3, f_4$. So, the next step, I think, would be to perform regression analysis on a subset of the original data set, containing only columns, corresponding to the extracted factors. Therefore, both goals will be achieved: performing EFA and regression.

• Sorry for my unclear question. Your understanding is correct. My following question is how I can use the re-patterned raw data to run regression analysis? Each extracted factor has been measured by 6-7 items/variables. How can I combined the 6-7 variables' raw data into ONE data set to represent ONE factor? – Ava XU Apr 16 '15 at 4:30
• Sorry...if the question is my pretty dummy level. :-) – Ava XU Apr 16 '15 at 4:35
• @AvaXU: No need to apologize - we are all here to learn :-). In regard to your subsequent question, I'm not sure that it's a good idea to answer it in comments. I will offer you a direction, but, if you need more detailed answer, I suggest you to either update your question by incorporating the second one, or by submitting a separate question. (to be continued) – Aleksandr Blekh Apr 16 '15 at 4:47
• @AvaXU: (cont'd) The direction: your second question implies that you're talking about latent variables, measured by a number of items (indicators) and, thus, latent variable model (LVM). Depending on the nature of your latent variables and indicators, a particular type of LVM should be used; see this presentation, pages 5-7. It is a large topic, so that is all I will say now. – Aleksandr Blekh Apr 16 '15 at 4:52
• Thanks for your direction. How can I give you a BIG HUG!? ;-) I need some time to digest the new concept. – Ava XU Apr 16 '15 at 4:57

2 major approaches on how to use Factor analysis with regression.

Approach1 conduct factor analysis and save scores for extracted factors. use these extracted factors as independent variable in your regression model. I have answered this already here

Approach2 conduct factor analysis to understand factor structure. Use summated scales to construct independent variable separately. quick tutorial on how to do this in SPSS