[Purpose:] We infer the posterior probability distribution functions (PDFs) and correlations of nine parameters characterizing the EOS of dense neutron-rich matter encapsulating a first-order hadron-quark phase transition from the radius data of canonical NSs reported by LIGO/VIRGO, NICER and Chandra Collaborations. We also infer the quark matter (QM) mass fraction and its radius in a 1.4 M$_{odot}$ NS and predict their values in more massive NSs. [Method:] Meta-modelings are used to generate both hadronic and QM EOSs in the Markov-Chain Monte Carlo sampling process within the Bayesian statistical framework. An explicitly isospin-dependent parametric EOS for the $npemu$ matter in NSs at $beta$ equilibrium is connected through the Maxwell construction to the QM EOS described by the constant speed of sound (CSS) model of Alford, Han and Prakash. [Results:] (1) The most probable values of the hadron-quark transition density $rho_t/rho_0$ and the relative energy density jump there $Deep/ep_t$ are $rho_t/rho_0=1.6^{+1.2}_{-0.4}$ and $Deep/ep_t=0.4^{+0.20}_{-0.15}$ at 68% confidence level, respectively. The corresponding probability distribution of QM fraction in a 1.4 M$_{odot}$ NS peaks around 0.9 in a 10 km sphere. Strongly correlated to the PDFs of $rho_t$ and $Deep/ep_t$, the PDF of the QM speed of sound squared $cQMsq/c^2$ peaks at $0.95^{+0.05}_{-0.35}$, and the total probability of being less than 1/3 is very small. (2) The correlations between PDFs of hadronic and QM EOS parameters are very weak. [Conclusions:] The available astrophysical data considered together with all known EOS constraints from theories and terrestrial nuclear experiments prefer the formation of a large volume of QM even in canonical NSs.