The NASA Kepler mission is designed to find planets through transits. Accurate and precise radii of the detected planets depend on knowing the radius of the host star accurately, which is difficult unless the temperature and luminosity of the star are known precisely. Kepler, however, has an asteroseismology programme that will provide seismic variables that can characterise stellar radii easily, accurately, and extremely precisely. In this paper we describe the Yale-Birmingham (YB) method to determine stellar radii using a combination of seismic and conventional variables, and analyse the effect of these variables on the result. We find that for main-sequence stars, a knowledge of the parallax is not important to get accurate radii using the YB method: we can get results to an accuracy and precision of better than a few percent if we know the effective temperature and the seismic parameters for these stars. Metallicity does not make much difference either. However, good estimates of the effective temperature and metallicity, along with those of the seismic parameters, are essential to determine radii of sub giants properly. On the other hand, for red giants we find that determining radii properly is not possible without a good estimate of the parallax. We find that the so called "surface term" in the seismic data has minimal effect on the inferred radii. Uncertainties in the convective mixing length can matter under some circumstances and can cause a systematic shift in the inferred radii. Blind tests with data simulated to match those expected from the asteroseismic Survey Phase of Kepler show that it will be possible to infer stellar radii successfully using our method.