ABSTRACT
Interest lies in modelling FFD as a function of the other variables. In this chapter, the authors focus on using multiple regression. In some of the panels the locally weighted regression fit differs considerably from the linear fit, confirming in some respect their suspicion about the form of the relationship between some pairs of variables. But for the moment, at least, the authors shall ignore this possible complication and find the results of the usual multiple regression model on the data. The most predictive variables appear to be SPR, RGF, and CAR, although since the regression coefficients and their standard errors are estimated conditionally on the other variables such an interpretation needs to be made with caution. Removing one of the nonsignificant explanatory variables and reestimating the remaining coefficients might lead to a different picture.
