In this paper, we introduce a new way to estimate the scaling parameter of a self-similar process by considering the maximum probability density function (pdf) of its increments. We prove this for H-self-similar processes in general and experimentally investigate it for turbulent velocity and temperature increments. We consider turbulent velocity database from an experimental homogeneous and nearly isotropic turbulent channel flow, and temperature data set obtained near the sidewall of a Rayleigh-Bénard convection cell, where the turbulent flow is driven by buoyancy. For the former database, it is found that the maximum value of increment pdf pmax(τ) is in a good agreement with lognormal distribution. We also obtain a scaling exponent α ≃ 0.37, which is consistent with the scaling exponent for the first-order structure function reported in other studies. For the latter one, we obtain a scaling exponent αθ ≃ 0.33. This index value is consistent with the Kolmogorov-Obukhov-Corrsin scaling for passive scalar turbulence but different from the scaling exponent of the first-order structure function that is found to be ζθ(1) ≃ 0.19, which is in favor of Bolgiano-Obukhov scaling. A possible explanation for these results is also given.