The distributions described in earlier chapters are based on certain assumptions about the processes underlying the variables. This chapter describes a few models that are flexible enough to be used in some situations where these assumptions do not hold.
- If the assumptions behind Poisson and binomial distributions do not hold, negative binomial and beta-binomial distributions have more flexible shapes.
- A Weibull distribution allows the hazard rate for a lifetime distribution to vary with age.
- Gamma distributions are often used as flexible models for variables whose values can be any positive value. In a similar way, Beta distributions are flexible models for variables whose values must lie between 0 and 1.
- Finally, normal distributions are symmetric distributions whose centre and spread can be adjusted to model many types of variable.