What would happen in a regression model if the intercept is suppressed and one of the dummy variables were dropped? Normally only one or the other is done to avoid the 'dummy variable trap'.

Answer 1

First Recognize That using dummy variables is usually an indication that we admit the model in not properly specified. That said, I will give you a simple. Suppose we were looking at Noon temperatures in New Cty. On Average April 15th and October 15th might have the Same Mean.

The True Relationship might be Temp =A*Sin((2pi(days past April 15th) +u. We could throw in a time trend if you wanted to test for global warming..

Instead of estimating this regression we regress temperature on a constant and dummy variables for each month other than April. We would expect the coefficients on December and January to be negative and June and July to be positive. October would be close to zero. This would probably give a pretty good predictive model. If we Suppress the constant and one dummy the other coefficients will suffer from specification bias as they "are being asked to explain more than they can)