Distributional
Assumption
The distributional
assumption means that assumptions are made on the distribution
of the outcome variable. With linear regression, the outcome
should have a normal distribution, or more specifically, the
residuals should have a normal
distribution. The logistic model makes the natural assumption
that the outcome follows a binomial distribution.
Survival
models usually assume proportional
hazards; i.e. that the relative risk is constant over
time.