The minimum margins required to compensate for random geometric uncertainties in the delivery of radiotherapy treatment were determined for a spherical Clinical Target Volume, using an analytic model for the cumulative dose. Margins were calculated such that the minimum dose in the target would be no less than 95% of the prescribed dose for 90% of the patients. The dose distribution model incorporated two Gaussians, and could accurately represent realistic dose profiles for various target sizes in lung and water. It was found that variations in target size and tissue density lead to significant changes in the minimum margin required for random errors. The random error margin increased with tissue density, and decreased with target size. The required margins were similar for dose distributions of spherical and cylindrical symmetry. Significant dose outside the spherical high dose region, as could result from multiple incident beams, lead to an increased margin for the larger targets. We could confirm that the previously proposed margin of 0.7 times the standard deviation of the random errors is safe for standard deviations up to 5 mm, except for very small targets in dense material.