The paper shows how to use the R package yuima available on CRAN for the simulation and the estimation of a general Lévy Continuous Autoregressive Moving Average (CARMA) model. The flexibility of the package is due to the fact that the user is allowed to choose several parametric Lévy distribution for the increments. Some numerical examples are given in order to explain the main classes and the corresponding methods implemented in yuima package for the CARMA model. of the Deutsche Mark/US Dollar daily exchange rate. Moreover [10] proposed the fractionally integrated CARMA model in order to capture the long range dependence usually observed in financial time series.The interest on the CARMA model is manifold since it can be used to model directly some given time series but it is also a main block for the construction of a more general process like the COGARCH(p,q) as in [5].The aim of this work is to develop in the yuima package a complete computational scheme for the simulation and the estimation of a general Lévy CARMA model. Based on our knowledge, the R packages available on CRAN deal only with CARMA(p,q) models driven by a standard Brownian Motion [30] or Gaussian CAR(p) models [33]. For example the ctarma package developed by [29] is an useful package for the simulation and the estimation of a CARMA(p,q) model driven by a Brownian Motion. Another package for continuous Autoregressive model is the cts developed by [33] which deals with a modified version of the CAR(p) model named CZAR(p) by [32]. Since the CAR(p) model is a special case of a CARMA(p,q), the ctarma package is a valid benchmark for the functions implemented in the yuima package and for this reason a direct comparison is given in this paper where a Gaussian CARMA(p,q) model is considered. Moreover, in the yuima package, once the estimation of the coefficients is done, it is possible to recover the underlying Lévy process from the observed data using the methodology in [7] and extended to the multivariate CARMA(p,q) by [8]. In this way we are able to simulate trajectories of a CARMA model without an explicit assumption on the distribution at time one of the underlying Lévy process. The outline of this paper is the following. In Sect. 2 we review the main results about the CARMA(p,q) process. In particular we focus the attention on the condition for the existence of the second order stationary solution of the CARMA process. In Sect. 3 we explain the estimation procedure implemented in the yuima package if the data are observed in equally space-time intervals. In Sect. 4 we describe the main classes and corresponding methods available in the yuima package for a CARMA model. We show how to use them for simulation and estimation of a Gaussian CARMA model and we conduct a comparison with the methods availables in the ctarma package. In Sect. 5 we present some numerical examples about the simulation and the estimation of Lévy CARMA models.