Dear all,

I am currently analyzing the effects several independent variables have on dependant variable (political party's position on immigration, to be more specific). As the cases (observations) here are political parties - the number of them is very limited (in fact, I have only 21 party with valid positioning on immigration). Thus, to make the results of OLS more accurate, I decided to 'expand' the pool of cases (observations). I have done this by including into the analysis parties of two different periods (2010 and 2014), thus increasing the total number of cases (observations) to 43.

After reviewing the analysis, my professor told that this may raise a problem as the assumption of independence of cases (observations) could have been broken. His argument is that there are parties, positions of which were measured in both 2010 and 2014 (let's say, party A had a position on immigration rated as 5 points (on 10-point scale) in 2010 and 6 points in 2014). Thus, his argument goes, that the cases (observations) are not independent.

However, after reviewing the result of Durbin-Watson test it showed that the assumption was not broken (DW value is 1,975).

Could you please suggest the way to progress further? Do I do nothing and say that the assumption was not broken (taking into account Durbin-Watson test). Or do I drop the cases (observations) that were measured in both time periods (but will that not decrease the accuracy of OLS as so few cases are analyzed)?

Thank you very much for your help!

Best wishes to everyone!