It is possible to run a spatial dynamic panel model by R. In doing so, you can estimate the speed of convergence for the sample over the period if there is no direct estimator available for dynamic panels with spatial dependencies. I propose you apply a two-step procedure, which involves first spatial filtering of the variables to isolate the spatial dependence, and apply standard GMM estimators for dynamic panels. The chances of misleading results would be slim by these measures. I am not too familiar with the use of Matlab in running panel data so can't say for sure as to it's suitability or otherwise.