Table 8 
(Temperature 90° F) 
  
  
  
  
SS df MS. 
Between! 89.30 | 3 | 2977 
Between-t 30.18 3 10.06 
Residual 102.26 9 11.36 
Total 221.74 15 
  
  
insignificant (below the 20 per cent level); but in this case the differences between 
the films fall between the 20 and 5 per cent points. Thus in the absence of other 
evidence there would be little justification in ranking the films in any particular 
order. 
The analysis of variance corresponding to storage at 120? F is given in 
Table 9. Here the time factor rises to a somewhat suspicious level, but remains 
  
  
  
  
Table 9 
(Temperature 120° F) 
SS... | dt! | MS 
Between-F 963.92 | 3 321.30 
Between-t | 45.92 3 15.31 
Residual | 76.89 9 | 8.54 
Total | 1086.75 15 
  
  
under the 5 per cent point. The differences between films, on the other hand, 
assume a high level of significance. Referring now to the original Table 6, 
we notice a wide gap between AB-284 and AP-311 on one hand, and AB-244 
and N-045 on the other. It can also be shown that N-045 is significantly better 
than AB-244 (t — 2.099, with 12 degrees of freedom, and, therefore P — .058). 
The difference between the partners in the other group gives + = 2.041, with 
12 degrees of freedom, and, hence, P = .064, which is very near the 5 per cent 
point. The superiority of the AB-284 over AP-311 cannot therefore be ruled out. 
Let us now compare the results which we obtained at the three temperatures. 
The variation of the film effect from insignificance in one case to a high 
level of significance in another, if itself is significant, would be indicative of a 
significant variation in the responses of the various film types to storage tempe- 
rature. Such an effect is usually referred to as an ’interaction between main 
effect’, in this case between the effect of the type of film (denoted by F) and the 
effect of temperature (denoted by 7). An interaction is usually denoted by a 
product of letters or numbers representing the interacting effects, thus, FT. 
The "t effect’, on the other hand, is not significant at any temperature level. 
We would not therefore expect a significant T't interaction. The present analysis 
gives no information on the Ft interaction. 
If however we assume that the original measurements were equally consis- 
tent, throughout the experiment and that the tabulated data are means of equal 
16