Fits a linear excess relative risk model by maximum likelihood, possibly including several variables and allowing for lagged exposures.