Many developmental processes in the life sciences, ecology and even in economics depend strongly on the environmental conditions occurring in a bounded time interval, the results occurring often far later. Examples are as diverse as plant phenology, grapewine maturation, diapause induction and so on. The method proposed here, aims at detecting quickly such effects. The basic idea is to regress the recorded results of a series of replications of the process against a function of an independent time series. This variable is defined on a set of periods of time, systematically scanned by varying their lower and upper bounds. In simple cases when this function is the integral and the effect strictly limited in a time window, the response model, under the form of correlation coefficients, is tractable and its shape is predictable. It is the same when the window is a bell-shaped function and can be fitted with other weighting functions as the Beta and the polynomials. The null hypothesis of absence of influence of any past interval is tested by Monte-Carlo simulation. The most likely window of influence is determined by the maximum correlation coefficient, and the bivariate confidence interval is estimated by bootstrap. The period found with a rectangular shaped window can be used as a starting point for more specific windows. This technique has the advantage of avoiding to split the climatic series into arbitrary slices, thus multiplying the predictors and complicating the models selection. It is closely linked to continuous lag distributed models with the simplification that the variable of interest is not explicitly time dependent. Examples are given for the prediction of aphids population dynamics, male morphs induction in aphids, and the phenology of the ash tree.

生命科学，生态学乃至经济学中的许多发展过程，在很大程度上取决于在有限的时间间隔内发生的环境条件，结果往往要晚得多。例子包括植物物候，葡萄成熟，滞育诱导等。这里提出的方法旨在快速检测这种影响。基本思想是使过程的一系列复制的记录结果相对于独立时间序列的函数进行回归。此变量是在一组时间段上定义的，可以通过更改其下限和上限来系统地对其进行扫描。在简单的情况下，当此函数是不可分割的并且在时间窗口内严格限制效果时，以相关系数的形式表示的响应模型是易处理的，并且其形状是可预测的。当窗口是钟形函数时，它是相同的，并且可以与其他加权函数（如Beta和多项式）拟合。通过蒙特卡洛模拟检验不存在任何过去间隔影响的零假设。影响的最可能窗口由最大相关系数确定，双变量置信区间由自举估计。带有矩形窗口的期间可以用作更特定窗口的起点。该技术的优点是避免将气候序列划分为任意片段，从而使预测变量相乘并使模型选择复杂化。它与连续滞后分布模型紧密相连，从而简化了所关注的变量与时间无关的情况。