INTRODUCTION
Approximation formulae for the chi-square distribution quantiles have been investigated in numerous papers beginning from Fisher [3] and Wilson and Hil- ferty [9]. Nowadays, very accurate approximation formulae are available (see e.g. Zar [10] and Johnson et al. [6], [7] or Ittrich et al. [5] and references therein). On the contrary, inequalities for the central chi-square quantiles rarely appear in the literature although they play an important role in some statistical considerations. Laurent and Massart [8] gave an exponential inequality for tails of the noncentral chi-square distrubution and used it to determine risk bounds for penalized estima- tor of the squared norm of a mean in a Gaussian linear model. This inequality is equivalent to some global upper bound for quantiles covering all values of param- eters involved. Brain and Mi [2] proved some upper and lower bounds which are expressed solely in terms of a number of degrees of freedom k and applied them to an interval estimation problem. Inglot and Ledwina [4] obtained lower bounds depending both on k and tail probabilities α and, employing them, described an asymptotic behaviour of the quantiles when k increases and simultaneously α tends