Accurately estimating earthquake magnitude within the first few seconds following an earthquake remains a major challenge in earthquake early warning (EEW) systems. This study explores and analyses various frequency and amplitude-dependent parameters to establish regression relationships between these parameters and earthquake magnitude. While commonly used parameters, such as characteristic period (τc) and peak displacement (Pd) are commonly used parameters for magnitude estimation, this work also incorporates additional amplitude-based duration-dependent metrics namely cumulative absolute velocity (CAV), cumulative absolute displacement (CAD), and cumulative absolute abasement (CAA). These parameters are computed over a 3-s time window to develop a regression model for magnitude estimation. We further examine the relationship between magnitude and purely amplitude-dependent parameters (Pd) following P-wave arrival. The analysis is based on 3902 strong-motion records from 327 earthquakes, recorded at 966 stations within the Japanese strong-motion network. Results suggest that amplitude-based parameters are more effective compared to frequency-dependent parameters in distinguishing between earthquakes with low and high peak ground acceleration (PGA), particularly within a 50 km hypocentre distance. Based on these observations, threshold values of CAV of 10 cm/s along, CAD of 0.2 cm, and CAA of 0.01 s/cm have been determined for each amplitude-dependent parameter, providing a useful reference for EEW applications.