试试chow test, 具体我不知道怎么运用在非线性模型里面。在线性模型里，就是比较pooled data 跟grouped data是不是有相同的参数。

在一个讲义看到下面一段话，希望对你有帮助！

Andrews and Fair (1988) extended the Chow test to apply in much more general settings –
• Nonlinear regression models with serially correlated and/or heteroskedastic errors (and weakly exogenous regressors).
• A wide class of nonlinear models (other than regression models) of dependent and heterogenous processes estimated by a variety of estimators.
These “Chow tests”, which include Wald-like tests, LM-like tests and LR-like tests depend on the correct specification of the breakpoint.