Hi Mohammed,

I am not exactly sure what you are trying to do.  Maybe the following questions will help.

Are you trying to create a single proxy variable that distinguishes just two classifications (i.e., either high or low agency conflict)? I think this is the case because it sounds like you are trying to segregate the analysis into two sub-samples.  Alternatively, a creating something like a continuum of scores from high to low could work for you if you are comfortable determining the cutoff score that separates the sub-samples?

Most of the time we use a proxy variable when we recognize the likely impact of something; however, we do not have a measurable variable that is directly influencing other variables.  Instead, it sounds to me like you are creating an index number, coefficient, or something else that captures multidimensional factors with your six different variables.

In the the of work we do, dummy variables and interaction terms are useful because their respective t-statistics provide a convenient way to test if the impact is statistically significant. Your work is different and so this may not apply.  It might help if you give us a better idea of the six variables.

I hope this helps and encourages other answers.

Rob

 Thanks Rob for your informative answer. 

Yes, as you said, what I am trying to do is to discriminate between companies with high and low level of agency conflicts. In this regard, I want to composite an index using 6 variables measuring agency conflicts to be used in distinguishing between companies. I am not sure if I can do that using a panel data because most of previous studies relied on cross-sectional data. do you have any idea if I can use panel data when I am using PCA to composite my index?

Thanks a lot :)