Open and Distance Learning

The standard values of the parameters in the model

Since it can be demonstrated that the û estimated parameters are asymptotically normally distributed [Bishop, Fienberg & Holland 1975], the standard values are also normally distributed. The ratio is
$$s (\hat{u}) = \frac{\hat{u}}{\sqrt{\hat{\sigma}^2(\hat{u})}}.$$

Calculating now for

^ uX(1)

in the numerical example, we obtain

s ( ^
u
 
X(1) 
) =
.77   ____
Ö.014
= 6.42 .

The parameter estimate is highly significant (p < .001), that is to say, the size of the effect, represented by û_X(1) , reflects the size of the marginal total (n_1. = 100), that for category 1 of variable X is the largest.


.