The test concerns the realization of phi-features on elements under the determiner node.
The short form includes the items with the highest phi coefficients and the highest item-total correlations among the 75 original items.
The class probabilities or relative frequencies are given by the also unknown vector phi (lines 9–14).
For this factor combination, the estimated precision coefficient phi exceeds the valid range for which the beta density can still be calculated.
Crucially, the inserted feature has to be checked before the inherent phi-features.
The preferred individual rule is, then, defined as that rule which maximizes the correlation coefficient phi for these individual predictions.
Thus, we obtain the following two (simpler) maximization goals: max pr(c|phi) wrt {phi} and max pr(x|{c, mu, sigma}) wrt {mu, sigma}.
Phi-features are the grammatical nominal features of person, number, and gender.
