行動計量学
Online ISSN : 1880-4705
Print ISSN : 0385-5481
ISSN-L : 0385-5481
最新号
選択された号の論文の12件中1~12を表示しています
原著
  • —死亡率上昇の意味すること—
    福地 幸文
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 77-88
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    This empirical study examines the changes in life insurance buying behavior caused by the Spanish influenza pandemic in Japan. We use panel data from the 47 Japanese prefectures during the years 1914-1922 to perform a regression analysis. By employing dummy variables, we demonstrate, that the Spanish influenza outbreak prompted life insurance buying behavior in 1919. In addition, a separate regression analysis using differences in mortality rates provides a glimpse at how social stability is the foundation of the life insurance business.

  • —ハードルモデルの消費者価格反応分析への応用—
    石橋 敬介, 伴 正隆, 尾崎 幸謙
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 89-97
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    We propose a model that consider the stage of purchase decision making by dividing it into deciding whether to purchase and deciding how many products to purchase. For this purpose, we used a hurdle model that combines a binomial logit model and a Poisson regression model. The hurdle model is suitable for analyzing zero-inflated data. In our analysis, we incorporated marketing variables in each model and compared which stage the marketing variables affect more. We compare several models using WAIC and show that the proposed model is superior. Results from the proposed model can be used for sales promotion activities targeted at individual consumers.

  • —どのような消費者レビューが「参考になった」を集めるのか?—
    金 勝鎮, 多田 伶, 勝又 壮太郎
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 99-109
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    This study proposes LDA with evaluation information. The authors focus on online consumer reviews with “helpfulness of review”. The purpose is to find specific topics that affect consumer's decision making from many consumer reviews. The authors exploit an evaluation score in addition to review text which is described in the natural language for simultaneously classifying topic model. The “helpfulness” scores are accessed by readers and are with integers of zero or more. The authors apply the suggested model in reviews for coffeemakers crawled from consumer's review website. The results show that the model distinguishes two topic groups with extremity which indicates that is helpful or not. Also, it shows that the two topic groups also have qualitative differences among firms (brands). The authors discuss the application of the model and findings for both review website managers and firm's marketers.

  • —情報処理プロセスに注目した実証的検討—
    多田 伶, 勝又 壮太郎
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 111-121
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    In this study, the moderating effect of consumers' information processing on post-purchase evaluations in an environment characterized by information overload was examined. Although it is important to understand how information processing during decision-making affects post-purchase evaluations, which include satisfaction with one's decision and repurchase intention, the influence of information processing on evaluations after purchasing is not well understood. Based on the unconscious thought study, two modes of thinking were compared: conscious thought and unconscious thought. To examine purchasing behavior regarding DVD players and window curtains, we conducted an online survey and performed path analyses. The results indicated that conscious thought had a negative impact on satisfaction with decision when consumers were confused by information overload. Moreover, it was suggested that unconscious thought had a positive impact on repurchase intentions when consumers felt confused by information overload. Finally, some implications and future research issues were addressed.

  • —マハラノビス距離を用いた機械学習による方法—
    東海林 和雄, 中村 亮介, 尾崎 幸謙
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 123-140
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    The purpose of this study is construction of the prediction model to discriminate incorrect accounting information. Two features of this research are to adopt methods of detecting auditing practices and to target for analysis that the accounting information which the sales are overestimated. Specifically, unlike in previous research, we approach to detect fraudulent means without uniform accounting phenomena for each fraudulent means. Furthermore, we applied accounting distortions and discomfort auditors feel as explanatory variables. This discomfort is measured by Mahalanobis distance. In the results of this analysis, the prediction model of machine learning that is adopted practical methods that detect incorrect accounting shows a high probability of fraud.

  • —Web調査用質問画面の場合—
    浅川 雅美, 岡野 雅雄, 林 英夫
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 141-152
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    This study examined the effect of the position of answer category “neither” of a rating scale (applicable/not applicable) of a self-administered questionnaire on the subjects' response rate, ease of answering, and misunderstanding of answer options in the case of a web survey. An eye tracking experiment was conducted on 56 university students. Fifty responses were analyzed and 6 were excluded due to insufficient data. The results showed that: (1) when “neither”is placed at the far right of the rating scale, the response rate is lower than when at the center; (2) when “neither” is placed at the far right, some participants mistakenly thought it displayed the most negative answer of “not applicable” because they did not observe the right side carefully; and (3) participants who correctly recognized the position of “neither” on the right had a higher fixation counts on the answer options and larger eye movements than those who did not recognize it.

資料
  • —階層線形モデルによるビッグデータの分析—
    菱山 完, 伊藤 徹郎, 岡田 謙介
    原稿種別: 資料
    2020 年 47 巻 2 号 p. 153-160
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    Iterative learning of simple assignments such as memorization of letters and words have shown to improve students' test scores of the learned assignments. Meanwhile, existing studies have mixed results as to whether simple repetitive learning improve the general academic performance or not. Using educational big data collected from more than four hundred thousand high-school students thorough crowd system, the current study investigated the effect of re-solving questions to the general academic achievement. Although descriptive statistics have not revealed steady tendencies, the results of hierarchical linear models that controlled for heterogeneity of schools, grades and individuals showed consistent positive effects of iterative learning towards general academic performance. The results suggest the importance of iterative learning, as well as controlling for the heterogeneity in large-scale educational dataset.

研究ノート
原著
  • 佐野 孝, 國土 将平
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 173-185
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    The aim of this study was to understand the movement difficulty in squat vault to examine the order and content of instructions given for the technique. Participants comprised of 241 children (116 boys and 125 girls) from 5th and 6th grades. Each participant performed squat vaults while their movements were observed and recorded from the left and front side. The movements were assessed using the observation criterion while the difficulty parameters and ability estimates were obtained by applying the partial credit model of Item Response Theory. From this data, the relation between ability estimation value and achievement of the technique, and order of the instruction contents were examined. The research conclusions were as follows:

    1) The evaluation based on the observational evaluation criteria reflects the difference in the degree of achievement of the skill.

    2) Considering the movement difficulty in squat vault, we can teach the skill in the following order: “weight transfer by arm support,”“hip rise by bouncing crossing,” “throwing arms forward and crouching forward,” “majestic jumping and stable landing,” and “turning back to backward rotation.”

  • —反応時間および識別率による流暢性の測定—
    西井 真祐子, 守口 剛
    原稿種別: 原著
    2020 年 47 巻 2 号 p. 187-197
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    The authors focus on consumers' reading direction for capturing product's visual information. There has been a great deal of research done, but these have discussed separately from product's image information and character information such as product's name. This study investigates the processes of the visual stimulus when product's information and product's image are processed simultaneously and how that affects product preference. Three experiment studies examined that effect of the direction of a product's cast shadow on processing fluency and product preference. Those revealed two points that (1) the consistency of direction between the product name written on the product package and the product's cast-shadow could enhance consumers' processing fluency, (2) the effect of the consistency of two directions on processing fluency may enhance product evaluation. Additionally, it is worth noting that we recruited several measuring methods for capturing “processing fluency” to reinforce the robustness of our findings.

資料
  • 財津 亘
    原稿種別: 資料
    2020 年 47 巻 2 号 p. 199-209
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    An integrated classification algorithm is a decision-making method that is not limited to a single classifier. It comprises multiple classifiers to maintain a high classification performance for various datasets. This study investigated the feasibility of an integrated classification algorithm for offender profiling. Offender profiling is the analysis of a crime scene using statistical and psychological methods to estimate information such as the age, job, and criminal record of the offender. In this study, the following 12 machine learning algorisms were used: decision tree (C5.0, CART by entropy or Gini), logistic regression analysis (LR), naïve bayes (NB), random forest (RF), bagging, boosting, support vector machine (SVM by radial basis function or polynomial), k-nearest neighbor (KNN), and neural network (NN). The results of the study showed that the classification performances of each algorithm varied for different objective variables of the dataset (e.g., criminal record, age, or job of offenders of residential burglar). However, the majority decisions made by a combination of three classifier algorithms (e.g., decision tree, LR, and NB) showed high classification performance regarding any dataset.

研究ノート
  • —アトリビュートの階層性がある場合の検討—
    丹 亮人, 岡田 謙介
    原稿種別: 研究ノート
    2020 年 47 巻 2 号 p. 211-225
    発行日: 2020年
    公開日: 2021/04/21
    ジャーナル フリー

    Cognitive diagnostic models (CDMs) are a class of statistical models that diagnose the mastery of respondents' cognitive traits, which are called attributes or skills. In the typical applications of CDMs, the Q-matrix, which represents which attributes are measured by each item, is specified by domain experts. In the case of dichotomous attributes, the impacts of Q-matrix misspecification on the classification accuracy have recently been studied; however, the case of polytomous attributes has not been reported. Therefore, in the present study, we examined how the difference between true and misspecified Q-matrix elements affects classification accuracy under four forms of attribute hierarchies. It was revealed that, in most conditions, larger difference between true and misspecified values resulted in lower classification accuracy. The impact of misspecification was the largest under the linear form of attribute hierarchy, which could be due to its smaller number of items that measure attribute levels. These results suggest that the number of items assigned to each attribute levels can be a key factor that affects the classification accuracy, especially when the degree of misspecification is large.

feedback
Top