I want to find the dominant one among two theories which give predictions about the relationships between one dependent variable( Y ) and some independent variables ( X ). The predictions of the two theories are mutually exclusive , i.e. if one theory predicts a positive relationship between Y and X1 , then the other theory predicts a negative relationship( or no relationship,only in one case ),etc.

I have conducted the appropriate regression analysis and based on the signs of the regression coefficients of the independent variables it may be seen which variables follow which theory . I want to test which theory is more dominant.

Considering the absolute values of the regression coefficients ( β ) , a weighted average β ( and also variances ), weighted by the factor [ ( 1 minus p-value)*100 ] is calculated for the two theories and then two sample one - tailed t - test ( with equal or unequal variances , as the case may be ) is used to test the difference in weighted average β for the two theories .

Is this procedure feasible or should I use other approaches ?

I have conducted the appropriate regression analysis and based on the signs of the regression coefficients of the independent variables it may be seen which variables follow which theory . I want to test which theory is more dominant.

Considering the absolute values of the regression coefficients ( β ) , a weighted average β ( and also variances ), weighted by the factor [ ( 1 minus p-value)*100 ] is calculated for the two theories and then two sample one - tailed t - test ( with equal or unequal variances , as the case may be ) is used to test the difference in weighted average β for the two theories .

**Note :**Since p-value is the actual level of significance , the factor ( 1 - p ) [ denoting the actual probability of accepting a true null hypothesis ] which assigns more weights to lower p - values would be more appropriate than assigning arbitrary weights . The multiplication by “100” is to dispense with fractional weights . The factor is to be rounded - off to the nearest whole number .Is this procedure feasible or should I use other approaches ?

Last edited: