56 
BULLETIN 1300, U. S. DEPARTMENT OF AGRICTTL.TUBE 
Better predictions can be made by taking more than one variable 
into account. Theoretically this could be done from the path- 
coefficient formulas, multiplying each coefficient by the ratio of the 
standard deviation of the dependent variable to that of the causal 
factor in question. Many of these formulas, however, involve the 
amount of breeding, a factor for which we do not have measurements. 
The best predictions can be obtained from the closely related mul- 
tiple-regression coefficients. The more important prediction formulas 
of this sort are given in Table 12. It is assumed that each variable 
is measured in terms of its standard deviation. Predictions in terms 
of actual units can be obtained as before by multiplying each term by 
the ratio of the standard deviation of the required variable to that of 
the correlated variable to which the term applies. The accuracy of 
49 \ 
/&■ 
^£{T<S&///G~ 4tfEST£&V ^tWSf&e #£/&/7T 
A37& /&?&- /&S& /&3S /<3&0 /&&S A9&? /&&f AP/0 AP/S" 
Fig. 30.— The actual western winter pack and that expected on the basis of the second and tlUrd preceding 
corn prices, and also that expected on the basis of the western summer live weight a year and a half 
before. There is only 36 per cent determination by these prices but 61 per cent determination by 
western summer live weight 
the fit during the period in question is indicated by the percentage 
determinations at the right. Figure 29 shows the fairly successful 
predictions of summer pack which could have been made on the basis 
of preceding corn prices (57 per cent determination) ; Figure 30 shows 
the less successful predictions of winter pack on the basis of corn 
prices (.36 per cent determination'). 
Table 13 gives a number of other multiple-regression formulas, 
which are not usable as prediction formulas since they involve con- 
temporary variables but are of interest in forming a judgment of the 
causal relations. 
The success of these formulas in predictions depends on the con- 
tinuance of the relations on which thev are based and also on the con- 
