I am just wondering where to find help for my question. Hopefully this is the right place with many smart people.

I am currently running some regressions with Stata. I started running normal OLS regressions. However, I just have dummy variables as independent variables and I am wondering if OLS is appropriate for this kind of model. Do you have any kind of help or comments on this problem?

Thanks a lot ]]>

I have extracted trends from two time series (one correspond to some climate parameters, the other one to chemical composition of tree cellulose) using loess functions (span 0.65, degree2). The result is that the trends (but not the residuals) of the data show almost parallel curves as function of time. Is there a way to calculate some correlation coeficient between the two trends (which are by nature highly autocorrelated)? if not, is there an objective way to demonstrate that the trends are "similar"?

Thanks a lot ]]>

At the moment I'm working on an analysis of wage inequality and have come across a considerable stumbling block. I am trying to compare the effects of level of education on logwages by groups. Hence, I am looking to add interaction terms into the equation; i.e. D_year8 x D_groupid. As I understand it, this should be somewhat simple i.e. the interaction variable for someone who completed year 8 and was in group 1 would be coded 1 (1x1=1); someone who completed year 8 but was in group 0 would be coded 0 (1x0=0) and so on until all interaction terms are within the equation. However as I add more interaction terms into the regression equation the coefficients for the interaction terms go from being quite plausible to downright ridiculous. Is there some fundamental part of this process I am missing? Any advice would be greatly appreciated :)

Cheers

Dave ]]>