We study a weighted least squares (WLS) estimator for Aalen's additive risk model which allows for a very flexible handling of covariates. We divide the follow-up period into intervals and assume a constant hazard rate in each interval. The model is motivated as a piecewise approximation of a hazard function composed of three parts: arbitrary nonparametric functions for some covariate effects, smoothly varying functions for others, and known (or constant) functions for yet others. The proposed estimator is an extension of the grouped data version of the Huffer-McKeague estimator [1]. Our estimator may also be regarded as a piecewise constant analog of the semiparametric estimates of McKeague & Sasieni [2], and Lin & Ying [3]. By using a fairly large number of intervals, we should get an essentially semiparametric model similar to the McKeague-Sasieni and Lin-Ying approaches. For our model, since the number of parameters is finite (although large), conventional approaches (such as maximum likelihood) are easy to formulate and implement. The approach is illustrated by simulations, and is applied to data from the Framingham heart study.