Chemical 1, 



N, points f, (X) = a, + a^X + ajX' . . apXP 



Chemical 2, 



N, points (X) = b, + b,X + b,X' . . bpXP 



Combine 1 and 2, 



N, + points fj (X) = c, + c^X + CjX^ . . CpXP 



Note: AW three regressions must be of the same 

 form; i.e., log, third degree polynomial, 

 and so forth. 



N, 



SSY, = 



SSY, = 



SSY = 



(y,i - f,(X|))' 



i = 1 



i = 1 



(N, + N,) 



E (yi - f3(Xi))^ 



i = 1 



SSY - SSY, - SSY, = difference in SSY 

 (N, + N, - P) - (N, - P) -(N, - P) = p 

 (SSY, + SSY,)/(N, + N, - P) = MSE(Y) 

 Diff SSY 



F(p, (N,+ N,- 2)) = 



MSE(Y) 



Figure 1.— Method used to calculate F values. 



The covariance analysis shows no significant difference be- 

 tween the fire-retarding effectiveness of any of the three 

 chemicals except for one parameter. There appears to be a 

 statistical difference between the regression for M-MAP and 

 S-MAP on excelsior weight loss. The greatest differences are 

 between data at the low treatment levels where small variations 

 in treatment amounts or fuel moisture content will sometimes 

 cause large-scale differences. (The curves show that fire retard- 

 ant effectiveness is very sensitive to small changes within the 

 low-treatment areas.) Examination of the M-MAP and S-MAP 

 regression data shows that their curves cross at about the 

 3-g/ft^ (929-cm^) treatment level. Apparent differences within 

 each separate regression for each chemical can be caused by 

 variation in fuel chemical composition, environmental condi- 

 tions, and retardant application. The apparent differences be- 

 tween the fire-retarding abilities of DAP, M-MAP, and S-MAP 

 are because of these variations in testing procedures; and the 

 apparent differences in total effectiveness are therefore not real 

 but because of experimental error. When the three chemicals, 

 two test parameters, and two fuel types are all combined, 

 analysis indicates no real significant differences exist (at the 

 0.05 level) between the fire- retarding abilities of DAP, 

 M-MAP, and S-MAP. 



Two methods were used to evaluate D-MAP, A-MAP, and 

 T-MAP. One method was to plot the individual points on the 

 curves for pooled DAP/M-MAP/S-MAP (PiOs curves) and 

 make general comparisons. The other method was to calculate 

 the flammability reduction percentage for each chemical and 

 compare it to the reduction calculated for the same level of 

 P^Os from the pooled DAP/M-MAP/S-MAP equation. Tables 

 12 through 14 show the percent reductions for the different 

 chemicals compared to reductions for pooled P2O5. The reduc- 

 tions compare closely for all chemicals with none varying more 

 than 0.01 from the pooled P2O5 regression. 



10 



