INRIA Saclay and University Paris-Sud, Orsay, France.
J Pharmacokinet Pharmacodyn. 2011 Dec;38(6):861-71. doi: 10.1007/s10928-011-9223-3. Epub 2011 Oct 26.
Visual Predictive Checks (VPC) are graphical tools to help decide whether a given model could have plausibly generated a given set of real data. Typically, time-course data is binned into time intervals, then statistics are calculated on the real data and data simulated from the model, and represented graphically for each interval. Poor selection of bins can easily lead to incorrect model diagnosis. We propose an automatic binning strategy that improves reliability of model diagnosis using VPC. It is implemented in version 4 of the MONOLIX: software.
可视预测检验（VPC）是一种图形工具，用于帮助判断给定的模型是否有可能产生给定的真实数据集。通常，时间序列数据被分为时间间隔，然后对真实数据和模型模拟数据进行统计计算，并为每个间隔进行图形表示。如果间隔选择不当，很容易导致模型诊断错误。我们提出了一种自动分箱策略，通过 VPC 提高模型诊断的可靠性。该策略已在 MONOLIX 软件的第 4 版中实现。
