The quantitywhere p is a proportion or probability. The term ‘logit’ was introduced by Berkson in 1944. Modelling variations in proportions directly is hampered by the need to ensure that estimated probabilities lie in the interval (0, 1). Since corresponding values for the logit lie in the unrestricted interval (−∞, ∞), models for proportions are usually constructed in terms of logits. See also probit.