Nonparametric estimation in change-point models (Q2366575): Difference between revisions

Assume that one observes data of the form \(y_ i=\theta+\varepsilon_ i\), where \(\theta=\theta_ 1\) and \(\theta=\theta_ 2\) for \(1\leq i\leq\tau_ 1\) and \(\tau_ 1<\tau_ 2\leq i\leq n\), respectively, and the \(\varepsilon_ 1,\varepsilon_ 2,\dots,\varepsilon_ n\) are i.i.d. with expection 0 and finite variance. For \(\tau_ 1<i<\tau_ 2\), \(y_ i\) is from a mixture distribution of the first and second population with specified mixing probabilities. The authors propose a least-squares approach to estimate the unknown change-points \(\tau_ 1\) and \(\tau_ 2\). Consistency is provided which in turn may be utilized to derive consistent and asymptotically normal estimators of \(\theta_ 1\) and \(\theta_ 2\).