Skip to main content

Table 8 Evolution of the standard deviation of residuals for various GRNN models predicting F a and F v in the RP–RF case

From: Deriving amplification factors from simple site parameters using generalized regression neural networks: implications for relevant site proxies

Number of parameters

Explanatory parameters

F a

F v

Standard deviation of residuals

Variance reduction

Standard deviation of residuals

Variance reduction

All (6)

Depth + f 0 + V sm + C v  + V s30 + V bedrock

0.0020

99.9%

0.00023

99.9%

3 (best triplet)

f 0 + C v  + V s30

0.0094

99.6%

0.0067

99.5%

2 (best pair)

f 0 + C v

0.0349

94.9%

0.0173

97.0%

2 (convenient pair)

f 0 + V s30

0.1083

50.7%

0.0515

73.2%

1 (best for Fa)

C v

0.0878

67.6%

0.0786

37.6%

1 (best for Fv)

f 0

0.1355

22.9%

0.0661

55.9%

1 (usual)

V s30

0.1419

15.4%

0.0752

42.9%

Initial σ

0.1543

0.0995