Emergence speed comparison by non-linear regression and approached by time-to-event models for censored data

Abstract

Abstract Determining the germination speed is essential in experiments in the field of seed technology, as it allows the performance evaluation of a seed lot and the creation of predictive models. To this end, the literature addresses several methods and indexes. The objective of this study was to compare the main methods of emergence speed analysis in seeds, namely the non-linear regression models and the Emergence Speed Index (ESI), with the time-to-event models. The research was conducted with peach palm seeds (Bactris gasipaes) that were measured for viability and vigour through daily evaluations for 4 months. Vigour was evaluated by the quantification of the seed emergence speed, which was performed in three ways: ESI, non-linear regression and non-linear regression considering germination as a time-to-event event. From the results obtained, we conclude that the ESI is not a good indicator to evaluate the emergence speed; the non-linear regression model underestimates the errors and, thus, increases the probability of misclassifying treatments; the time-to-event model is more reliable in classifying treatments according to the emergence speed.