Cambridge University Press, Weed Science, 3(70), p. 259-271, 2022
DOI: 10.1017/wsc.2022.8
Full text: Unavailable
AbstractGermination and emergence assays represent the most notable examples of time-to-event data in agriculture and related disciplines. In spite of the peculiar characteristics of this type of data, there has been little effort to establish a specific and comprehensive framework for their analyses. Indeed, a brief survey of the literature shows that germination and emergence data, along with other phenological measurements such as flowering time, have been analyzed through myriad approaches, giving rise to confusion and uncertainty among scientists and practitioners as to what may represent the best statistical practice. This lack of coherence in statistical approach may reduce the efficiency of research, while making the communication of results and the cross-study comparisons extremely challenging. Here, we attempt to provide a coherent framework and protocol for the analyses of germination/emergence and other time-to-event data in weed science and related disciplines, together with a software implementation in the form of a new R package. We propose a similar approach to biological assays in ecotoxicology, based on: (1) fitting a time-to-event model to describe the whole time course of events; (2) comparing time-to-event curves across experimental treatments, and (3) deriving further information from the fitted model to better focus on some traits of interest. The most appropriate methods to accomplish this procedure were carefully selected from the framework of survival analysis and related sources and were modified to comply with the specific needs of weed, seed, and plant sciences. Finally, they were implemented in the new R package drcte. In this article, we describe the procedure and its limitations by way of providing examples of several types of germination/emergence assays. We highlight that our proposed procedure can also serve as the first step of data analyses, with its output subsequently submitted to traditional or meta-analytic approaches.