Shiga University Repository
Not a member yet
15011 research outputs found
Sort by
キョウツウ カリキュラム ニ オケル ショウガイ ノ アル コドモ ヘノ シエン ノ コウゾウ アメリカ ガッシュウコク ノ シュウガクシュウ スタンダード モデル カリキュラム ニ チュウモク シテ
Get PDFdepartmental bulletin pape
Generalizing the Permanent Income Hypothesis with Homothetic Robust Epstein-Zin Utility
Get PDFThis study assumes homothetic robust Epstein-Zin utility and a market model in which the state vector follows a general Markovian diffusion process. The study derives an optimal consumption rule as a generalized permanent income hypothesis (PIH) and demonstrates that the marginal propensity to consume out of the total wealth can be decomposed into five terms, four of which are functions of the state vector. This indicates that the generalized PIH does not ensure the stability of consumption as implied by Friedman’s PIH. Additionally, the study derives the conditional expected growth rate of consumption and demonstrates that it can be decomposed into four terms. The first term is the difference between the interest and discount rates. The second and third terms are both positive and interpreted as the “effect of precautionary savings on risk” and “effect of precautionary savings on ambiguity,” respectively. The fourth term is interpreted as the “effect of timing of resolution of uncertainty.” The stability of the expected growth rate of consumption is not ensured either, because all these terms are functions of the state vector.technical repor
The Reality of the Long-Unfinished-Railroad Company Continuously Exploited by “Fake Enterprisers” : Early-Stage History of the Yashima Railroad Company, a Subsidiary of a Model Village at the Foot of Mount Chokai in the 1920s
Get PDFThe Yashima Railroad was planned in 1920 primarily to transport timber from the foot of Mount Chokai in Akita Prefecture, with official authorization for track construction granted in 1921. However, despite partial construction of the track, the railroad never began operation, and the government revoked its authorization in 1935. In the years that followed, the company pursued administrative litigation and various petitions, including those to the General Headquarters of the Supreme Commander for the Allied Powers (GHQ) during the postwar period. Nevertheless, all efforts ended in vain with the Yashima Railroad becoming the most tragic privately operated, small-scale railroad in Japan.
When a flood in 1926 washed away the track under construction, the original construction plan was on the verge of abandonment. Nishitakizawa Village , commended by the government as a model village, decided under the strong leadership of its mayor to rally the community and reconstruct the railroad track. A few knowledgeable individuals questioned the decision to integrate the railroad company into the village as a subsidiary, but their dissent was silenced to maintain village unity. Meanwhile, a variety of propositions were made, including offers by contractors to advance construction costs and by specialized firms to lease rails, vehicles, and other equipment. Misled by these seemingly attractive offers, the village lost the opportunity to fundamentally reassess its construction policy.
As a result, many of the village residents became shareholders of the railroad company, and the village’s cooperative association shouldered the shortfall in construction costs. The eventual bankruptcy of the company led to the immediate decline of the community and the impoverishment of the majority of its residents. When the author visited what was once a model village to speak with its residents, many of the elderly with painful memories of the lingering effects of the experience, either directly or indirectly, adamantly refused to discuss it. To them the “kamakeyasu” railroad remained a taboo topic, with “kamakeyasu” meaning “bankruptcy” in the Akita dialect.
To uncover the mystery behind the company’s “sealed history,” this study examines how, during its early years, the company was continuously exploited by numerous “swindling enterprisers,” leading to soaring initial expenses, the misallocation of funds to cover unnecessary costs, and the depletion its capital. These factors severely deteriorated the company’s financial condition, resulting in an excessive burden placed on the villagers who eventually took over its management.departmental bulletin pape
フクザツ ナ コウゾウ オ モツ コウジゲン データ ノ ヘンスウ スクリーニング ホウ ニ カンスル ケンキュウ
Get PDF滋賀大学博士（データサイエンス）Regression analysis is widely used to model relationships between variables and is applied in applications such as future prediction and measuring the effect of variables. For example, to understand the regularity of product sales, sales are treated as the response variable, while product attributes (for example, price, color, size, etc.) serve as predictors. The predictive accuracy and interpretability of the model are considerably affected by the variables included in the model. Appropriate variable selection enhances model accuracy and interpretability, resulting in improved decision-making and outcomes across various fields.
Several methods have been devised for variable selection. The stepwise method sequentially adds or removes variables to improve the model, whereas regularization methods with sparsity-inducing penalties reduce some coefficients to zero by applying a penalty based on the size of regression coefficients. However, for high-dimensional data, where the number of variables considerably exceeds the sample size, the computational cost becomes substantial because of the large number of variables and tuning parameters requiring evaluation. Moreover, these methods often fail to consistently identify the truly important variables.
Numerous screening methods have been proposed to address these problems. These methods evaluate the correlation between each predictor and the response, selecting variables with strong correlations. By evaluating predictors individually, screening methods are computationally efficient. Additionally, many screening methods exhibit the sure screening property, which ensures that the probability that the selected variables contain truly important variables converges to one as the sample size increases. However, their performance degrades in the presence of multicollinearity, and they typically do not account for interactions among predictors. Real data often exhibit complex structures, such as multicollinearity and interactions, which, if not addressed, can reduce model accuracy and interpretability.
This thesis proposes novel approaches to handle multicollinearity and interactions, respectively. First, we introduce a screening method that improves variable selection in the presence of multicollinearity by modifying an existing factor analysis-based approach. The existing method uses factor analysis on predictors to transform the data and reduce multicollinearity. However, it can unintentionally discard too much information about important predictors, leading to lower variable selection performance. The proposed method prevents this problem by truncating a portion of specific common factors obtained by factor analysis during the transformation process, with the truncation determined objectively using a BIC-type criterion. After appropriately transforming to remove multicollinearity, we select predictors that have large absolute values of correlation with the response. A theoretical model based on the eigenvalue distribution demonstrates the improved performance of the proposed method. Simulated and real data analyses confirm its effectiveness.
Second, we propose an interaction screening method that enhances variable selection performance for multi-class classification problems by modifying an existing approach based on Kendall’s rank correlation coefficient. The proposed method focused on an approach that does not consider assumptions about the existence of main effects and directly determines the importance of the interactions because, in some cases, only interactions are meaningful in the real world. The existing method directly computing the importance of interactions tends to underestimate the importance of interactions related to minor classes in datasets with highly imbalanced class labels because its importance score depends on correlations calculated from all of the sample and on the ratios of class sample sizes. To address this issue, the proposed method incorporates correlations computed using data from each class and the average ratios of sample sizes across multiple classes. The proposed interaction importance scores more accurately reflect the information from the minor classes. Additionally, we show that the proposed method satisfies certain theoretical properties. One is the sure screening property. Another is the ranking consistency property, which guarantees that interactions truly related to the response obtain higher importance scores than unrelated interactions. Using the ranking consistency property, we can identify the truly important set of interactions by selecting interactions in order of their importance scores if the number of important interactions is known and the sample size is sufficiently large. Simulated and real data analyses reveal that the method effectively identifies important interactions even in imbalanced datasets.doctoral thesi
A Contrastive Study of Personal Pronoun Usages: An Analysis of Multilingual Folktales
Get PDFIf we look at the languages of the world, we find that some languages, such as English and Chinese, make extensive use of personal pronouns, while others, such as Japanese, use them much less. In this study, we used parallel texts from several languages to investigate the differences in personal pronoun usage. We argue that these differences are not only due to grammatical rules but also reflect cultural rules and constraints.
In this study, we investigated the use of personal pronouns in world fairy tales (Snow White, Hansel and Gretel, and the Three Little Pigs; the original text was a bilingual English- Japanese text), specifically in the Amele language of Papua New Guinea and Tok Pisin (an English-based creole).
In (1) below, English is a language that uses personal pronouns extensively, and Tok Pisin also uses personal pronouns extensively because some parts of its grammar are based on English grammar.
(1) Snow White: 3rd person “she”
● English: as she sat sewing at her window,
● Tok Pisin: em sundaun na sanap long window bilong em,
In (1) Tok Pisin, the personal pronoun “em” corresponds to the English “she/her.” However, the third-person singular personal pronoun “em” is the same for both males and females, so it can refer to either “he” or “she.”
(2) Snow White: 2nd person “you”
● English: No, you idiot, she's asleep.
● Japanese: Iya, baka danaa, nemu-tteiru dake- dayo.
In (2), the English personal pronouns “you” and “she” do not appear in the corresponding Japanese text. Generally, Japanese is a language that tends to avoid personal pronouns.
(3) Snow White: Person name
● English: What Snow White did not know was that the cottage belonged to seven dwarfs.
● Amele: Snow White uqa bedoor uju yo uju nacbataak agena.
In (3), we look at an example from the Amele language spoken in Papua New Guinea. In English, the third-person singular is expressed by the name Snow White, but in Amele, after introducing the name Snow White, the thirdperson singular personal pronoun “uqa” is used, creating a parallel with “Snow White uqa”.
In this study, we analyzed multilingual texts and found that the usage of pronouns differs between English and other languages. Notably, personal pronouns are frequently used in languages other than Japanese, which we investigated this time.departmental bulletin pape
