Dow Divisor Impact on OLS Regression

I'm doing an econometric project for university based on analysis of capital assett pricing model using the DJIA over a three year period (12th Oct 2007 - 12th Oct 2010). However, when it comes to plugging the data into Eviews to run some OLS regression involving the return to the index compared to the return to individual assets, I'm unsure as to whether or not to used the published values of the index, the simple summation of the stock values, a basic price-weighted index or an index based on a fixed/variable value for the divisor. I'm maintaining the components as the ones found at the end of my test period (e.g. no citigroup) by creating an artifical index meaning should I create artifical daily index values using the summation of my index's assets divided by the real world value of the divisor given on each perticular day?
However, I can imagine this will mess with my results as I assume the value of the dow divisor is adjusted to compensate for events related to companies not present in my index.

Thanks in advance for any help.

Matt

Also, sorry if this was posted in the wrong area.