行動観察における観察頻度の決め方—サンプリング法予備研究

Determination of frequency in behavior observations: Pilot study of sampling method

要旨：行動観察における頻度予測のためのサンプリング法に関する予備研究を行った．5分等間隔で行動観察を行ったときの出現率と，観察間隔10〜60分でシミュレーションした出現率の間に差があるか比較した．その結果，母集団とした5分等間隔サンプリング(データ数145)による観察対象事象の出現率と，10〜60分間隔サンプリング(データ数73〜13)による標本の出現率の間には，50標本中49標本に有意差は認められなかった．つまり，30〜60分の観察頻度でも5分間隔の行動観察と同程度の予測ができる可能性が示され，今後検討を進めていく必要性を見出した．

　Objective. The objective of this study was to investigate the extent to which observation frequency could be reduced to obtain a necessary mean in behavioral observations.

　Method. The behaviors of 2 subjects were observed for 12 hours with a sampling frequency of a 5minute interval. The data obtained was used as the population, and data obtained from a Bernoulli trial, in which the sampling frequency was gradually reduced, was used as the sample for simulation. The incidence of events in the sample was compared with the incidence of the population event. The required sampling frequency was determined from an estimated precision formula.

　Results. There was no significant difference in events between the present study and the incidence of events observed in the 5-minute interval sampling (number = 145) and in the 10〜60-minute interval sampling (number = 73〜13).

　Conclusion. We found that 30〜60-minute interval sampling led appropriately to the result: a 5minute interval sampling of observation. It is necessary that more research be done to explore the frequency of behavior observation.