Let $\{X_i\}_{i=1}^n$ be a random sample, $Y_i=\delta(X_i)$ for some function $\delta(\cdot)$, and $Q(\tau)$ be the population quantile function of $Y_i$. We can estimate $Q(\cdot)$ by the empirical quantile function $\widehat Q(\tau)$ based on the random sample $\{Y_i\}_{i=1}^n$. Under certain regularity conditions, we have \begin{align*} \sup_{\tau\in(0,1)}|\widehat Q(\tau)-Q(\tau)|=o_P(1). \end{align*}

Now suppose we do not know the function $\delta(\cdot)$. Instead, we have a uniform convergent estimator $\widetilde\delta(\cdot)$ for $\delta(\cdot)$, that is, \begin{align*} \sup_{x\in\mathcal{X}}|\widetilde\delta(x)-\delta(x)|=o_P(1). \end{align*}

Given the pseudo sample $\{\widetilde Y_i\}_{i=1}^n$, $\widetilde Y_i=\widetilde \delta(X_i)$, and let $\widetilde Q(\tau)$ be the empirical quantile function based on the pseudo sample, do we have \begin{align*} \sup_{\tau\in(0,1)}|\widetilde Q(\tau)-Q(\tau)|=o_P(1)? \end{align*} How to show this and what conditions are needed?

Thanks a lot for your help.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.