= Maximum-likelihood fitted probability
A <maximum-likelihood fitted probability> evaluates a model <probability> at a <maximum-likelihood estimate>: $\widehat p(x)=p_{\widehat\theta}(x)$. For independent <Bernoulli trials> with common success <probability> $p$, the fitted <probability> is $\widehat p=K/n$, the observed success fraction. In <logistic regression> it instead varies with the predictor through $\widehat p(x)=(1+e^{-x^T\widehat\beta})^{-1}$. It is an estimated <event> <probability>, rather than the <probability> that the fitted model is correct or a <p-value>.
Back to article page