1: \begin{abstract}

2: The delta method is a popular and elementary tool for deriving

3: limiting distributions of transformed statistics, while applications

4: of asymptotic distributions do not allow one to obtain desirable

5: accuracy of approximation for tail probabilities. The large and

6: moderate deviation theory can achieve this goal. Motivated by the

7: delta method in weak convergence, a general delta method in large

8: deviations is proposed. The new method can be widely applied to

9: driving the moderate deviations of estimators and is illustrated by

10: examples including the Wilcoxon statistic, the Kaplan--Meier

11: estimator, the empirical quantile processes and the empirical copula

12: function. We also improve the existing moderate deviations results

13: for $M$-estimators and $L$-statistics by the new method. Some

14: applications of moderate deviations to statistical hypothesis

15: testing are provided.

16: \end{abstract}