When the sample sizes are very large.

When the population variances are equal.

When the data are heavily skewed or violate the assumption of normality.

When the researcher wants to prove a causal link.