Abstract
It was shown in Groeneboom (1983) that the least concave majorant of one-sided Brownian motion without drift can be characterized by a jump process with independent increments, which is the inverse of the process of slopes of the least concave majorant. This result can be used to prove the result in Sparre Andersen (1954) that the number of vertices of the smallest concave majorant of the empirical distribution function of a sample of size $n$ from the uniform distribution on $[0,1]$ is asymptotically normal, with an asymptotic expectation and variance which are both of order $\log(n)$. A similar (Markovian) inverse jump process was introduced in Groeneboom (1989), in an analysis of the least concave majorant of two-sided Brownian motion with a parabolic drift. This process is quite different from the process for one-sided Brownian motion without drift: the number of vertices in a (corresponding slopes) interval has an expectation proportional to the length of the interval and the variance of the number of vertices in such an interval is about half the size of the expectation, if the length of the interval tends to infinity. We prove an asymptotic normality result for the number of vertices in an increasing interval, which translates into a corresponding result for the least concave majorant of an empirical distribution function of a sample of size $n$, generated by a strictly concave distribution function. In this case the number of vertices is of order cube root $n$ and the variance is again about half the size of the asymptotic expectation. As a side result we obtain some interesting relations between the first moments of the number of vertices, the square of the location of the maximum of Brownian motion minus a parabola, the value of the maximum itself, the squared slope of the least concave majorant at zero, and the value of the least concave majorant at zero. An erratum is available in EJP volume 18 paper 46 .
Highlights
Submitted to EJP on March 25, 2013, final version accepted on March 31, 2013
It has been brought to my attention by Svante Janson ([2]) that formula (2.9) of part (iii) of Lemma 2.1 in [1] is scaled incorrectly
Part (iv), which is deduced from (iii), is still correct, as is the derivation of (iv) on the same page, since the factor 2−2/3 arises from differentiating the characteristic function of V (0) in (iii) twice with respect to t at t = 0
Summary
It has been brought to my attention by Svante Janson ([2]) that formula (2.9) of part (iii) of Lemma 2.1 in [1] is scaled incorrectly.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.