125
from 3.0 to 6.0 in. This equation had an R2 of 0.933 for 134 number of
cases. Predictions were generally within 10 percent of actual Eg
values. Only 22 of the 134 total observations had more than 20 percent
predictive error. It must also be noted that the use of this equation
requires knowledge of the other layer moduli; namely E2 and E4. If
any of these is off, then the E3 value predicted from Equation 4.14
would probably be in error. Thus, the reliability of the E3 prediction
is dependent on the accurate estimates of the other layer moduli.
Equation 4.15, which predicts E3 from D4-D?, is a relatively sim
plified equation. Table 4.7 shows the high degree of accuracy of this
equation to predict E Correlation between predicted and actual E
values resulted in an R2 of 0.991. Extrapolation of this equation to
predict outside the stipulated range resulted in errors as high as +60
percent. However, it will be shown later that this equation format,
using field measured Oynaflect deflections, is applicable to a wider
range of variables than those listed under Case 3.
4.4.1.4 Subgrade Modulus, E Equation 4.16, derived to predict
the subgrade modulus, holds for a wide range of E4 (0.35 to 200.0
ksi). This equation has an R2 of 0.984 for 266 number of cases.
Because of the wide range of E4, the equation tends to underpredict
slightly for intermediate E4 values. Also, predictive errors greater
than +60 percent were found to be typical for the extreme values of E4
(0.35 and 200.0 ksi). Therefore, those data were deleted and the
remaining data set regressed to obtain Equation 4.17 with an R2 of 0.992
and N equal to 193. Equation 4.17 holds for E4 values between 10.0 and
50.0 ksi. Only 4 out of 193 E4 predictions fell above +15 percent.