How can a P-Value be obtained for an analysis of Eta association? Is there a reference for this?
Given that your previous title mentioned nominal and numerical, I assume you are talking about the effect size measure eta-squared. Eta-squared is the sample proportion of variance explained in a numerical variable by a categorical predictor variable. It is a biased estimate of the population variance explained (for a relatively unbiased estimate, I've often seen omega-squared recommended).
A p-value could be obtained by running a one-way between subjects ANOVA.