We present results for the quasar clustering on large (greater than or similar to 10 h-1 Mpc) scales, using a sample of about 750 quasars. Our statistic is based on xi(r)BAR = (3/r3) integral-r/0 xi(x)x2 dx where xi is the two-point correlation function of quasars (unless otherwise stated, the comoving separation r is calculated by assuming a flat universe with LAMBDA = 0). Significant clustering is detected at r less than or similar to 30 h-1 Mpc. For 10 less-than-or-equal-to r less-than-or-equal-to 50 h-1 Mpc, xi(r)BAR is well described by a power law xi(r)BAR = A/r(gamma) with gamma = 1.8 and A = 74 +/- 10 [corresponding to a correlation length r0 = (6.6 +/- 0.5) h-1 Mpc]. The best fit of xiBAR to the power spectrum P(k) is-proportional-to k[1 + (klambda(p)/2pi))2.4] gives lambda(p) approximately 200 h-1 Mpc, showing that P(k) may bend to the n approximately 1 spectrum around this wavelength. This result is in good agreement with recent observations of galaxy clustering at low redshifts, and suggests that the prediction, given by some models of galaxy formation, that quasars trace the same large-scale structure as galaxies and clusters do is consistent with the observed quasar correlation. If xiBAR is-proportional-to (1 + z)-2, as predicted by an Einstein-de Sitter dust universe with linear density perturbations, the amplitude of the quasar correlation function would give a correlation length r0 approximately 18 h-1 Mpc at z = 0, suggesting that high-redshift quasars may form at high-density peaks in the initial density field. For 10 less than or similar to r less than or similar to 30 h-1 Mpc, our analysis shows a marginal (2 sigma) detection of evolution in the quasar correlation function, with A(z < 1.5)/A(z > 1.5) approximately 2.0. This evolution is best described by a model where xiBAR is-proportional-to (1 + z)-2. But the statistic is still too poor to constrain evolutionary models.
