The study on scaling the scrape-off layer (SOL) power width (λq) is crucial for deepening the understanding of the SOL particle and heat transports. Due to the sparse distribution of the divertor Langmuir probes and the erosion of probe tips during the long-pulse high-performance operations on EAST, the estimation of SOL particle flux width (λjs, used to approximate λq) from the measured ion saturation current density profile (js) usually has relatively large uncertainty. This paper introduces a maximum a posteriori estimation method based on the Bayes' theorem to reduce the fitting uncertainty for λjs (the fitting accuracy increases by 33% in terms of mean absolute error compared with the traditional ordinary least squares estimation). With the new estimation method and the FreeGS equilibrium code, the databases in Liu et al (2024 Nucl. Fusion 64 026002) are updated, which are further used to scale λjs. Compared with the old λjs scalings for the L-mode and H-mode databases in deuterium and helium plasmas, the updated λjs scalings show better regression quality with similar results. The deuterium and helium databases for L-mode and H-mode plasmas can be combined to get a unified scaling, , where is the averaged SOL connection length, is the fraction of Greenwald density, is the poloidal beta, PSOL is the power crossing the last closed flux surface (LCFS), is the surface area of the LCFS, and Z is the charge number. The unified scaling reveals that: (i) λjs has a strong scaling dependence on the SOL connection length suggesting the missing scaling dependence on the machine size for the Eich scaling; (ii) the helium λjs is slightly larger than the deuterium λjs. Furthermore, the scalings for integrated particle flux width are also given in this paper.