Can a Linear ARX Neural Network be created but just changing the nonlinear function of the hidden layer of a NARX network to a linear function?

I have created, trained and tested a NARX network using narxnet(). I am trying to run the exact "same" network but as a linear ARX network. Can I do so by changing the nonlinear function in the hidden layer? if so, how can I change it? Will changing to a linear function change any of the internal traning concepts?

I cannot find like a arxnet() command to keep my current data format and avoid reformating to use something like arx().

Thanks.