Professional Documents
Culture Documents
Rosenblueth 1975 Point Estimates For Probability Moments
Rosenblueth 1975 Point Estimates For Probability Moments
Rosenblueth 1975 Point Estimates For Probability Moments
USA
Vol. 72, No. 10, pp. 3812-3814, October 1975
Mathematics
ences. vy [10]
y =I y+ y-
TWO-POINT ESTIMATES
Let X and Y be real random variables and Y = Y(X) a well- V _I1 Y+ -
Y-I [11]
X2 X3
4p I+p
a
X2
I
4
I+
4 _
x_ xI
X2
° 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Xi XI
FIG. 1. Concentrations of probability density function when Y
= Y(X1, X2)-
FIG. 2. Concentrations of probability density function when Y
= Y(X1, X2, X3).
2P+(x+ -X) = 3ax4 [14]
Eq. 14 stems from the condition E[(X -)4] = 3oX when X Proceeding in like manner with n = 3, we obtain the con-
is gaussian. Solution of Eqs. 12-14 gives P+ = 1A, P = %, and centrations in Fig. 2, where the center of the hexahedron is
at X1, X2, X3, and
x+= X + qx/3
We may proceed analogously with estimates at a larger P+++ = P___ = (1/8)(1 + P12 + Paw + P-31)
number of points. Results can be obtained from the table in
ref. 3 (4). The larger the number of points, the more robust P++_= P__+ = (1/8)(1 + P12 - ;3 -P3)
will be the procedure relative to the irregularities in Y pro- P+-+ = P-l = (1/8)(1- P12 - P21 + P31)
vided X is gaussian.
We now seek a procedure using a three-point estimate ro- P+_ = P a+ = (1/8)(1- p12 + Pt P31)
Downloaded from https://www.pnas.org by 189.147.94.22 on November 20, 2023 from IP address 189.147.94.22.
It follows from Eq. 18 that E[(X -X)3 = X3VX(3 + VI). 1. Benjamin, J. R. & Cornell, C. A. (1970) Probability, Statistics,
Hence, vx = VX(3 + V2] = 0.608. According to Eqs. 6-8, and Decision for Civil Engineers (McGraw-Hill Book Co.,
P+ = 0.35457, . With k = 3 the exact answer is, from New York).
Eqs. 18 and 19, Y = 1.125, Vy = 0.651. Eq. 1 yields the 2. Rosenblueth, E. (1974) "Aproximaciones de segundos momen-
same value of Y and Vy = 0.693 while, according to Eqs. 9 tos en probabilidades," Boletin del Instituto Mexicano de Pla-
and 11, Y = 1.120, Vy 0.543. neacibn y Operacibn de Sistemas 26, November-December,
1-12.
Consider now Y = XjX2, X1_= X2 = 1, ax, = qx2 = 0.5, p 3. Salzer, H. E., Zucker, R. & Capuano, R. (1952) "Table of the
= 0 (2). According to Fig. 1, Y = 1.5625, Vy = 1.639. Eqs. zeros and weight factors of the first twenty hermite polynomi-
16 and 17 yield the same estimate for ? and Vy = 1.300. als," Journal of Research National Bureau of Standards,
With the same data but oxi = OX2 = 0.1, Fig. 1 gives Y = RP2294, 48,111-116.
1.0201 and Vy = 0.282. These coincide to the last significant 4. Abramowitz, N. & Stegun, L. A. (1968) in Handbook of Mathe-
digit with results of applying Eqs. 16 and 17. matical Functions (Dover Publications, Inc., New York), p.
924.
5. Rosenblueth, E. (1975) Optimum resource allocation to re-
I am indebted to Niels C. Lind for his helpful criticisms.
search, to be presented at Fifth European Symposium on
Earthquake Engineering, Istanbul, September 1975.
Downloaded from https://www.pnas.org by 189.147.94.22 on November 20, 2023 from IP address 189.147.94.22.