|
|||||
|
|
Robust confidence intervals for trend estimation in meta-analysis with publication bias |
|
| Authors: | H. Lu P. Yin R.X. Yue |
| Affiliation: | 1. School of Mathematics and Statistics, Nanjing University of Information Science and Technology, Nanjing, People's Republic of China;2. Department of Biostatistics, University of Liverpool, Liverpool, UK;3. College of Mathematics and Sciences, Shanghai Normal University, Division of Scientific Computation of E-Institute of Shanghai Universities and Scientific Computing Key Laboratory of Shanghai Universities, People's Republic of China |
| Abstract: | Confidence interval (CI) is very useful for trend estimation in meta-analysis. It provides a type of interval estimate of the regression slope as well as an indicator of the reliability of the estimate. Thus a precise calculation of confidence interval at an expected level is important. It is always difficult to explicitly quantify the CIs when there is publication bias in meta-analysis. Various CIs have been proposed, including the most widely used DerSimonian–Laird CI and the recently proposed Henmi–Copas CI. The latter provides a robust solution when there are non-ignorable missing data due to publication bias. In this paper we extended the idea into meta-analysis for trend estimation. We applied the method in different scenarios and showed that this type of CI is more robust than the others. |
| Keywords: | meta-analysis publication bias random-effects model robust confidence interval trend estimation |