The relationship between the variations of shear-wave splitting parameters and the preparation or occurrence of earthquakes has been one of the most interesting scientific problems in shear-wave research. It has not been possible to establish a physical correlation between the variation of the shear-wave time-delays and the change of stress during an earthquake, which can be confirmed by the vast majority of observations owing to the large scatter in the complete set of time-delay measurements of shear-waves. Therefore, it is necessary to analyze the cause of the scatter of the time-delays between fast and slow shear-waves in near-field seismic records and to identify a method to reduce the scatter of time-delays. Scatter in the measurement of the fast and slow shear waves may arise from the anisotropic structure and complexity of the medium under the observation station. It may also originate from the data processing procedure. To determine the source of the scatter in the measurement of shear-waves time-delays, we first need to identify the scatter, and distinguish its effects from the actual three-component seismic shear-wave records or measurements of the shear-wave parameters. This study is based on single station near-field seismic data from the KHZ station in northern South Island, New Zealand, from January 2013 to January 2017. It uses the visualization method to process the shear-wave data and measure the shear-wave splitting parameter of the seismic event in the shear-wave window of the KHZ station. It assumes that the physical properties and structure of the medium under the observation station are basically stable. The frequency filtering and distance normalization factors used to process the shear-wave, which lead to scatter in the measurements of the shear-wave time-delays, are compared one by one. In addition, the “false scatter” phenomenon, which may be caused by the multiple splits resulting from the anisotropic stratification of the medium under the observation station, is analyzed. Conclusions and suggestions for correcting the scatter in the measurement of shear-waves are proposed as follow: (1) the selection of the filter window directly affects whether the filtered observation data retain the waveform and amplitude characteristics of the shear-wave. If the main frequency of the shear-wave in the observation window is used as the main reference for the selection of the filter band, the obtained measurement of the shear-wave splitting parameter is more stable when the majority of the main frequency component is retained and there is less interference from the other frequencies. (2) The use of the whole source distance in the path normalization of shear-wave is another prominent factor in the scatter of the shear-wave time-delays. This is unreasonable in many cases because the source distance is often greater than the actual length that the split shear-wave propagates in an anisotropic medium. This “excessive” normalization causes all the time-delays between fast and slow shear-waves in the observed window to decrease with increasing source distance. (3) The actual path length of the seismic shear-wave through the anisotropic medium, that is, the “effective source distance,” should be the path length used in normalizing the time-delay. The depth of the interface where shear-wave splitting occurs should be determined by correcting the “excessive” normalization of the shear wave time-delays when there is no significant seismic activity near the observation station. The lower interface depth of the anisotropic medium under the KHZ station, by the near-field shear-wave data, is 37 km. (4) After correcting the scatter of the time-delays, the shear-wave multiple splitting phenomenon caused by the anisotropic stratification of the medium under the observation station is identified. From the “shear wave pairs” in the multiple splitting of the shear-wave data, it can be determined that there are other layers in the anisotropic medium below the KHZ station, and the corresponding depths of the interface are 15 and 27 km, respectively. (5) Normalized time-delay results from the KHZ station shows that before the 7.8 magnitude earthquake that occurred on November 13, 2016 at a distance of 50 km, the time-delays became increasingly large for a few months, and quickly decreased after the earthquake. It is a strong indicator of the stress accumulated and released in the anisotropic medium near the observation station before and after larger earthquakes. This aspect is expected to become an observable precursor of earthquakes.