Page 120 - Basic Statistics

115

t testing criteria:

To a certain significance level testing , H0 is accepted if

t-actual   t/2;n-p-1 , or

Pr = [ P ( t  -t actual) + P ( t  t actual) ]  , otherwise H0 is rejected.

In this partial test, it should be noted that if the results of testing a

variable coefficient is statistically significant at a model that involves all the
independent variables, it is not necessarily significant to the model with only a

subset of the independent variable.

Partial t-tests for linear regression models in worked example 5.4, are

shown in Table 5.12 below. Just as the previous F-test results that only constant

and coefficient 5 are significantly different from zero.

Table 5.12 Coefficients

Model Unstandardized Coefficients Standardized t Sig.
Coefficients

B Std. Error Beta

(Constant) -6,512 ,934 -6,976 ,000
X1 1,999 2,573 ,223 ,777 ,449
X2 -3,675 2,774 -,371 -1,325 ,204

X3 2,524 6,347 ,105 ,398 ,696

X4 5,158 3,660 ,366 1,409 ,178
X5 14,401 4,856 ,671 2,966 ,009
a. Dependent Variable: Y

5.2.7 SEQUENTIAL F-TEST

Sequential test is used to study the contribution of a variable in the

model containing the preceding regressor variables. The order of entry, then,

can have a profound effect on the results. If regressor 4 is adjusted for 1, 2, and

3, it is quite possible that its contribution to SS(Reg) will be quite different than

~~* CHAPTER 5 THE MULTIPLE LINEAR REGRESSION MODEL *~~