1: \begin{abstract}

2:     In this paper, we study a nonlinear cointegration-type model of the form

3:     \(Z_t = f_0(X_t) + W_t\) where \(f_0\) is a monotone function and \(X_t\)

4:     is a Harris recurrent Markov chain. We use a nonparametric Least Square

5:     Estimator to locally estimate \(f_0\), and under mild conditions, we show

6:     its strong consistency and obtain its rate of convergence. New results

7:     (of the Glivenko-Cantelli type) for localized null recurrent Markov chains

8:     are also proved.

9:   \end{abstract}

10: