Estimating regression equations with common explanatory variables but unequal numbers of observations