We develop a general procedure to fit a splicing model to censored and/or truncated data. We propose to combine the flexibility of the mixed Erlang distribution to model the body of the distribution with the Pareto distribution to provide a suitable fit for the tail.