We consider three different methods of generating non-Gaussian Markov processes with given probability density functions and exponential correlation functions. All models are based on stochastic differential equations. A number of analytically treatable examples are considered. The results obtained can be used in different areas such as telecommunications and neurobiology.