(Sub- and supermartingale inequalities) (a) Let \(\{X_1, \dots, X_n\}\) be a sub-martingale. If \(\lambda \ge 0\), show that
\(\lambda P\{\max_{1 \le i \le n} X_i \ge \lambda\} \le \int_{\{\max_{1 \le i \le n} X_i \ge \lambda\}} X_n dP \le E(X_n^+) \).
(b) Let \(\{X_1, \dots, X_n\}\) be a submartingale. If \(\lambda \ge 0\), show that
\(\lambda P\{\max_{1 \le i \le n} X_i \ge \lambda\} \le E(X_1) - \int_{\{\max_{1 \le i \le n} X_i < \lambda\}} X_n dP
\le E(X_1) + E(X_n^-) \).
(Apply 6.7.3, with \(T = \min\{i: X_i \ge \lambda\}\); \(T = n\) if all \(X_i < \lambda\).
(c) If \(\{X_1, X_2, \dots\}\) is a submartingale and \(\lambda \ge 0\), show that
\(\lambda P\{\sup_n X_n > \lambda\} \le \sup_n E(X_n^+) \);
if \(\{X_1, X_2, \dots\}\) is a supermartingale and \(\lambda \ge 0\), show that
\(\lambda P\{\sup_n X_n > \lambda\} \le E(X_1) + \sup_n E(X_n^-) \).