Page 425 - Practical Design Ships and Floating Structures

400

TABLE 1.
GOODNESS OF FIT FOR REGRESSION MODEL I

complex correlation Multivariate multivariate estimator of
coefficient determination of adjusted standardization
efficient determination residual
R2 = SSR / SST coefficient
P(I e: 1 <2)
,=I R2=1- SSEIdJE
a SSTIdfT
0 59520
lll_-" ..I_x__ -I -- 0.35426 -I-_- 1 1 0.33781 -. 0.96914
~
-
3.2.3 Significance test of regression parameters
The significance of the population linear correlation in the regression model doesn't mean each
independent variable has a significant influence over dependent variables, so the significance test for
each regression parameter is necessary. However, each regression parameter in the equation of
stepwise regression algorithm is got on the condition that each factor significance is considered to be
beyond certain significance. Thus, the significance test for each regression parameter may be ignored.

3.2.4 Calculation of Beta coefficients

As well known, Beta coefficient is an ideal indication to weigh the influence of each argument over
function,

v ,=I
where Beta coefficient b, measures the x variable of the j,, with other x variables kept fixed, and
reflects the influence of alternation of one standard unit over y. Substituting b: for b, can measure
the influence among variables under the condition of no-dimension number. The calculation results of
Beta coefficient are shown in table 2:

TABLE 2. Beta coefficient of influence factors in model I:
4 Lh Dh 4,
-0.34523 0.26248 -0.27714 -0.56427

4 EXPERIMENTAL CORRECTION OF MATHEMATICAL MODELS FOR LINE HEAT
FORMING

In the calculation of local contraction values, the curve plate is assumed as a theoretical hull without
thickness. However, the practice of real ship plates processed by line heat forming shows that the
influence of the thickness of plates shouldn't be ignore. So a series of experiments for regular plates of
vaned thickness are carriedout. In those experiments, quantitative analyses of the influence of plate