Authors:Fei Gu; Hao Wu Abstract: In canonical correlation analysis (CCA), the substantive interpretations of the canonical variates are of primary interest to the applied researchers. However, there are two different interpretive approaches used by different researchers—the weight-based approach and the loading-based approach, of which the latter is favored by the majority of researchers in practice. For those who choose the loading-based approach and apply CCA simultaneously to multiple samples, they may wish to test the invariance of the canonical loadings. In this paper, three covariance structure models are defined for CCA. In particular, the first model (i.e., the CCA-W model) corresponds directly with regular CCA, including the canonical correlations and canonical weights as the parameters, while the third model (i.e., the CCA-L model) is in alignment with the loading-based interpretive approach, including the canonical correlations and canonical loadings as the parameters. The CCA-L model is further extended to the unrestricted and restricted SCCA-L models, of which the latter allows one to test the invariance of the canonical loadings. A real example drawn from the sociological literature is provided to illustrate the restrictive SCCA-L model, and some strategies to calculate good starting values for the restrictive SCCA-L model are discussed. PubDate: 2017-10-25 DOI: 10.1007/s41237-017-0042-8

Authors:Augustus Jayaraj; Oluseun Odumade; Stephen Sedory; Sarjinder Singh Abstract: In this paper, we propose a new weighted squared distance while minimizing a distance between the true proportions and the observed proportions of (Yes, Yes), (Yes, No), (No, Yes) and (No, No) answers in the setup of Odumade and Singh (Commun Stat Theory Methods 38: 439–446, 2009) model. The resultant estimator is shown to be an unbiased estimator of the proportion of the sensitive attribute of interest in a population and has smaller variance than the estimator of Odumade and Singh (Commun Stat Theory Methods 38: 439–446, 2009) with the same protection of the respondents. A real data application is also considered. PubDate: 2017-10-24 DOI: 10.1007/s41237-017-0041-9

Authors:Shuzo Abe; Yoshiyuki Okuse Abstract: When two alternatives are compared, the one that becomes the focus of comparison has a relative advantage (resp., disadvantage) when two alternatives share positive (resp., negative) unique features. This “direction of comparison effect” (DCE) has been found to manifest clearly when three conditions are met: when information is retrieved from memory rather than being externally presented, when a consumer has low product involvement but a high need for cognition, and when the attribute set size is large. The authors argue that at least two additional conditions must be met when applying the DCE in marketing: (1) the presence of unique features of market offerings and (2) the absence of situational factors such as commitment to an alternative. Experimental data on choice behavior regarding business hotels provide support for these two conditions. Also, in discussing the results, it has become clear that unnoticed DCE bias should be carefully canceled out in surveys by designing questionnaires to ensure that each alternative randomly becomes the focal alternative. PubDate: 2017-10-11 DOI: 10.1007/s41237-017-0039-3

Authors:Hanako Ohmura Abstract: While egotropic evaluations of economic policy have long been analyzed as reflecting the absolute individual income situation per se, economic assessments should also comprise relative perceptions of others’ status; in other words, voters’ perceptions of economic inequality matter. First, using an internet survey, we examined how perceptions of economic disparity affect income tax rate preferences founded on absolute income. We then tested whether egotropic and sociotropic evaluations separately affect voters’ retrospection on government, or if the easier egotropic evaluation functions as an information cue for the more difficult sociotropic evaluation. The results demonstrated the following: (1) those who perceive economic inequality to exist tend to be punitive toward voters in other income brackets; (2) this punitive tendency is stronger among higher and middle earners; and (3) for the middle earners, absolute income status and for the higher earners, perception of economic inequality mediates the effect of sociotropic evaluations and ultimately influences government approval. PubDate: 2017-10-11 DOI: 10.1007/s41237-017-0040-x

Authors:Tomohito Okabe; Daisuke Nogiwa Abstract: Political party preference is a crucial element in the analysis of economics and political science. However, it is often difficult to investigate the dynamic properties of the individual partisanship due to inaccessibility to panel data. This study proposes a Bayesian approach for estimating Markov dynamics of individual-level partisanship with repeated cross-section data in which the history of respondents’ choice of favored party cannot be observed. The proposed approach identifies individual heterogeneities that affect transitional patterns of partisanship, and replicates the dynamic patterns of individual partisan mobility. Using the proposed method with American survey data, the study shows that age, education and race significantly influence partisan dynamics among Americans over a three-decade period from 1972. PubDate: 2017-09-19 DOI: 10.1007/s41237-017-0034-8

Authors:Masahiro Morii; Takayuki Sakagami; Shinya Masuda; Shigetaka Okubo; Yuki Tamari Abstract: Internet surveys are currently used in many academic and marketing research fields. However, the results for these surveys occasionally show traces of response bias. In our study, we analyzed how response bias appears in lengthy preference judgments. 1042 respondents participated in lengthy sequential preference judgments. Three stimuli series were used: scene pictures, Attneave nonsense shapes, and point-symmetric figures. One hundred stimuli were selected for each series and individually displayed on a computer screen, with presentation order randomized for each respondent. Respondents were then asked to rate their degree of preference for each stimulus. Mean preference scores increased over the first 10–20 trials, then, gradually decreased from the middle to the last trial. Furthermore, participants tended to produce the maximum and minimum score during early trials. These results demonstrated that response bias can be a function of presentation order. PubDate: 2017-09-18 DOI: 10.1007/s41237-017-0036-6

Authors:Koji E. Kosugi Abstract: Using Bayesian inference, this study aims to estimate the magnitude of the cognitive load when a person perceives asymmetric social relations. Some empirical evidence relating to balance theory has shown that a balanced state is comparatively easier to memorize than an unbalanced one. In this study, since a balanced state is defined by structural complexity, an experimental hypothesis was set whereby asymmetric social relationships have different difficulty levels depending on structural complexity. The balanced state of an asymmetric relation as structural difficulty is formally derived from the eigenvalue structure of a Hermitian matrix. Asymmetric triadic relations are modeled as featuring three kinds of structural difficulties according to the eigenvalue decomposition of the Hermitian matrix and pattern-specific difficulties. The differences among the structural difficulties were not sufficiently significant to exceed pattern-specific difficulties, but the Bayes factor of the informational hypothesis of this research yielded positive effects. PubDate: 2017-09-14 DOI: 10.1007/s41237-017-0033-9

Authors:Marcus Selart; Svein Tvedt Johansen Abstract: This study tested a model of the relationship between work performance, employee personality, and perceived career development. The model hypothesized that employee personality (extraversion) modifies the relationship between work characteristics (task, diversity, work feedback) and perceived career development. Results revealed that work characteristics (task diversity, work feedback) was positively related to perceived career opportunities and that extraversion positively moderated the relation between work characteristics (task diversity, work feedback) and perceived career opportunities. Issues related to the common method bias are discussed (see e.g. Podsakoff et al. in J Appl Psychol 88(5):879–903, 2003). PubDate: 2017-09-12 DOI: 10.1007/s41237-017-0035-7

Authors:Haruhiko Ogasawara Abstract: The 3-parameter logistic (3PL) model including guessing parameters is one of the popular models in item response theory. While the guessing parameters in the fixed-effects 3PL model with non-stochastic abilities have been believed to have identification, some counter examples with new ones given in this paper are currently available. In this paper, the concept of degeneracy for nested models is introduced. Some degenerate cases in the counter examples are shown to have model identification for guessing parameters, which are further shown to have model unidentification in a more degenerate model. Similar results in the fixed-effects 4-parameter logistic model are also derived. PubDate: 2017-08-28 DOI: 10.1007/s41237-017-0032-x

Authors:Haiyan Liu; Zhiyong Zhang Abstract: Measurement error in continuous, normally distributed data is well known in the literature. Measurement error in a binary outcome variable, however, remains under-studied. Misclassification is the error in categorical data in which the observed category is different from the underlying one. In this study, we show, through a Monte Carlo simulation study, that there are non-ignorable biases in parameter estimates if the misclassification is ignored. To deal with the influence of the misclassification, we introduce a model with false-positive and false-negative misclassification parameters. Such a model can not only estimate the underlying association between the dependent and independent variables, but it also provides information on the extent of the misclassification. To estimate the model, a maximum likelihood estimation method based on a Fisher scoring algorithm is utilized. A simulation study is conducted to evaluate the performance, and a real data example is given to demonstrate the usefulness of the new model. An R package is developed to aid the application of the model. PubDate: 2017-08-23 DOI: 10.1007/s41237-017-0031-y

Authors:Tsuyoshi Hatori; Satoshi Fujii; Kazuhisa Takemura Abstract: The objective of this study was to examine the effects of people’s prior choice making on their weights of decision attributes. According to theories on post-decision processes, preferences are likely to be reconstructed in line with a preceding choice. The present study hypothesized that preferences are constructed so that superior (inferior) attributes of a chosen alternative are weighted higher (lower) than before. This hypothesis was tested in a field survey of university students in Japan. The data supported the hypothesis that the weights of the superior (inferior) attributes of the chosen alternative increased (decreased) as time passed since the decision was made. This result suggests the validity of the causal relationship in which choice shapes preference, which is the reverse of what is generally assumed in decision research. PubDate: 2017-07-27 DOI: 10.1007/s41237-017-0028-6

Authors:Shinya Masuda; Takayuki Sakagami; Hideaki Kawabata; Nobuhiko Kijima; Takahiro Hoshino Abstract: Some studies have shown low proportions of Japanese people rating their lives as happy. However, most surveys in Japan revealed that response distributions for subjective well-being items indicated excessive blips in middle categories. In this study, participants responded to four items that were designed to detect careless or unmotivated respondents; these items instructed respondents to choose a certain response category. Of the study’s respondents, about 40% failed to follow instructions for at least one of the four items. Besides, they often chose middle categories for well-being items. When those who did not follow instructions were excluded from the complete sample, middle categories showed low or no blips, and the proportion of respondents with high rating scores increased. Thus, this study suggested that one reason for reportedly low proportions of happy people in Japan was respondents who were not diligent in reading items. PubDate: 2017-07-01 DOI: 10.1007/s41237-017-0026-8

Authors:Maarten Marsman; Lourens Waldorp; Gunter Maris Abstract: Large-scale prediction problems are often plagued by correlated predictor variables and missing observations. We consider prediction settings in which logistic regression models are used and propose a novel approach to make accurate predictions even when predictor variables are highly correlated and only partly observed. Our approach comprises three steps: first, to overcome the collinearity issue, we propose to model the joint distribution of the outcome variable and the predictor variables using the Ising network model. Second, to render the application of Ising networks feasible, we use a latent variable representation to apply a low-rank approximation to the network’s connectivity matrix. Finally, we propose an approximation to the latent variable distribution that is used in the representation to handle missing observations. We demonstrate our approach with numerical illustrations. PubDate: 2017-06-14 DOI: 10.1007/s41237-017-0024-x

Authors:Patrick Blöbaum; Takashi Washio; Shohei Shimizu Abstract: It is generally difficult to make any statements about the expected prediction error in an univariate setting without further knowledge about how the data were generated. Recent work showed that knowledge about the real underlying causal structure of a data generation process has implications for various machine learning settings. Assuming an additive noise and an independence between data generating mechanism and its input, we draw a novel connection between the intrinsic causal relationship of two variables and the expected prediction error. We formulate the theorem that the expected error of the true data generating function as prediction model is generally smaller when the effect is predicted from its cause and, on the contrary, greater when the cause is predicted from its effect. The theorem implies an asymmetry in the error depending on the prediction direction. This is further corroborated with empirical evaluations in artificial and real-world data sets. PubDate: 2017-04-10 DOI: 10.1007/s41237-017-0022-z

Authors:Agostino Tarsitano; Ilaria Lucrezia Amerise Abstract: In cases where only linear relationships are suspected, Pearson’s correlation is generally used to measure the strength of the association between variables. It is well-known, however, that when a non-linear or non-linearizable connection exists, the use of Pearson’s coefficient on original values can be deceiving. On the other hand, rank correlations should perform satisfactorily because of their properties and versatility. There are many coefficients of rank correlation, from simple ones to complicated definitions invoking one or more special transformations. Each of these methods is sensitive to a different feature of dependence between variables The purpose of this article is to find a coefficient, if one exists, that tends to be different from zero at least in a meaningful way more often than others when the relationship between two rankings is of a non linear type. In this regard, we analyze the behavior of a few well-known rank correlation coefficients by focusing on some frequently encountered non-linear patterns. We conclude that a reasonably robust answer to the special needs arising from non-linear relationships could be given by a variant of the Fisher–Yates coefficient, which has a more marked tendency to reject the hypothesis of independence between pairs of rankings connected by several forms of non-linear interaction. PubDate: 2017-03-27 DOI: 10.1007/s41237-017-0020-1

Authors:Tomoya Okubo; Shin-ichi Mayekawa Abstract: One approach to smoothing empirical score distributions is to apply a probability distribution. However, the score distributions sometimes form a ‘toothed’ shape, meaning that the probability distribution does not always fit well. The normal distribution can be applied for the fitting; however, this has a symmetric shape and a domain ranging from minus infinity to plus infinity. Polynomial log-linear models are often used for fitting frequency distributions, especially those with complicated shapes, but these do not provide information on the frequency distribution. In this study, we propose the maximum likelihood model for approximating a discrete distribution. Our approach is based on the mixed-multivariate beta distribution, which can cover asymmetric score distributions with known upper and lower bounds. Further, a simulation study is conducted to compare the results from the application of the mixed-multivariate beta distribution, mixed multivariate normal distribution, and two polynomial log-linear models. Results from a real data analysis are also given in this research. PubDate: 2017-03-20 DOI: 10.1007/s41237-017-0019-7