where ε j∼N (0, ) such that ε j≥1−α −X jβ , j= 1,…, n . To gain more precise confidence intervals, this study uses the parametric bootstrap for regression process to derive the confidence intervals for the estimates of parameters (β , ). For an overall picture and the details of the estimation algorithm, readers are encouraged to refer to Simar and Wilson [32] and [33].
We perform a simulation test with a total of 3000 experimental observations to confirm the fitness of the truncated regression model, which is specified as follows: