Although statisticians often assume normally distributed errors, there are important processes for which the error distribution has a heavy tail. A well-known heavy-tailed distribution is the t distribution, but the t distribution is unsuitable for some applications because it does not have finite moments (means, variance,...) for small parameter values.