APPENDIX B 



B-3 



99.6 

 99.6 



99. 8 

 99. e 



COUPON RANGE IN 1000 PSI 



PANEL RANGE IN 1000 PSI 



Fig. B-l. Tensile Strength, 

 Coupon Range for Mat (Ml) 

 and Cloth Faced Mat (M4) 

 Laminates - All Angles Included 



Fig. B-2. Tensile Strength, 

 Panel Ranges for Mat (Ml) 

 and Cloth Faced Mat (M4) 

 Laminates - All Angles Included 



Thus in the body of this report, different values are shown for the two materials and a 

 different pattern is shown with angle for each material. Finally, the fabricators must 

 be distinguished from each other since one was able to produce material with a 3. 6 ksi 

 consistently greater tensile strength than the other three fabricators. 



The best estimate we can get of the error standard deviation among panels made at 

 different times will be obtained by pooling mean squares for MF, MT, FT and MFT 

 if these are all of comparable magnitude. Such is the case here. The combined Mean 

 Square, with 24 degrees of freedom, is 10. 6. This Mean Square estimates 3o p + °a 

 + o c /2. We take as an estimate of panel variance, then, 10.6/3 or 3.53, remembering 

 that this does include some within-panel variability (in the present case about 0. 33 

 from 0.98/3). The corresponding standard deviation is 1.9. 



Having checked on the near normality of the data (by the ecd plots of the specimens, 

 of the panels, and of the discrepancies computed in the analysis of variance) it 

 appeared safe to use published tolerance factors to estimate ranges below means; 

 (called L.T.R. for Lower Tolerance Ranges, in the tables of this report). These 

 values, subtracted from the averages estimated in each cell of the tables, give the 

 Lower Tolerance Limits (L.T.L.) shown. All tolerance limits are computed at 

 the 95 per cent level of confidence. They are also the limits that are expected to 

 be lower than 95 per cent of the population that corresponds. The multipliers used 

 were published by G. Lieberman, in the Journal of Industrial Quality Control for 

 April 1958. For the present example, at 95 per cent confidence, and for 95 per cent 

 coverage, with 24 degrees of freedom, the factor is found to be 2. 30 and the lower 

 tolerance range is therefore 2. 30 x 1. 9 or 4. 4. 



Similar operations were carried through for fourteen properties and for three groups 

 of materials. With only a few exceptions, the data were readily interpretable. The 

 exceptions were in the data on compressive Poisson's ratio and in some of the per 

 cent water absorption figures. The irregularity of the data on these two properties, 

 at least for some of the materials, greatly diminishes the value of these sets. 



