This study examined whether responding to many items increased middle category response. In two internet-based experiments, respondents were randomly assigned to one of the four versions of the questionnaire. In Study 1, the respondents completed a Big Five questionnaire that had varied item arrangements. The results showed that the endorsement of the middle category increased for items placed later in the questionnaire for all versions. In Study 2, the Big Five questionnaires differed in that all items were administered at once or were facially divided into two scales. The results indicated that less middle categories were selected in the first 5 items of the second half scale than in the last 5 items of the first half scale for the divided formats, but the endorsement rates increased for items placed later toward the end of both half scales. An increasing middle category response was also found for another scale. Therefore, this phenomenon seems reliable.
Self-reported surveys in media usage behavior have been conducted by many companies and institutions. However, whether such surveys accurately reflect the actual usage has rarely been discussed. In this study, we revealed the difference between self-reported survey and automatically collected behavior log for smartphone usage duration, and investigated the relationship of individual attributes to the difference. Though past studies focused on measures (frequency versus duration) or question formats (open-ended versus closed), we successfully reduced the difference by adopting questioning methodologies, i.e., recall aids and timetable. We found that; (1) smartphone usage duration tends to be underreported, (2) when reported results highly deviate from the log, applying recall aids and timetable questionnaire reduces the difference, and (3) the difference is affected by age, gender, full-time job, and leisure time.
The item count technique, which is an indirect questioning method, aims to yield more valid results in social surveys than the direct questioning method by eliminating the effects of social desirability on responses. The technique requires respondents to only answer the number of applicable items on an item list. However, the mean responses of the technique are often smaller than those of the questions that require respondents to select “applies” or “does not apply” for each item. This underreporting tendency of respondents often prevents the technique from yielding valid results. Tsuchiya and Hirai (2010) proposed an elaborate item count technique for reducing underreporting. Based on this technique, respondents are asked to reply both the number of applicable and non-applicable items. In this study, the data used by Tsuchiya and Hirai (2010) is re-analyzed from the viewpoint of response latency. The analysis reveals that underreporting of the item count technique is remarkable in short-time respondents. The analysis also elucidates that the elaborate item count technique demands more response time than the direct “applies/does not apply” questions and the technique successfully suppresses underreporting even in short-time respondents. The redundancy of the elaborate item count technique is considered to be a main cause of suppressing underreporting by forcing respondents to more thoroughly examine their responses.
The purpose of this study was to examine how the interpretability of a purchased product changes by mindset manipulations based on the Mental Simulation (MS) or Construal Level Theory (CLT). While the manipulation of MS associates the targeted products, that of the CLT does not. We investigated whether this difference affects the accuracy of the survey. In this experiment, participants were randomly assigned to four groups (photo-simulation manipulation (MS) vs. how-manipulation (MS) vs. low-level manipulation (CLT) vs. non-manipulation); they completed each task and conjoint measurement. The result of the experiment showed that photo-simulation manipulation and how-manipulation better contributed to the estimation of the real purchased product than non-manipulation did. However, this was not the case for low-level manipulation.
The receiver operating characteristic (ROC) curve is a currently well-developed statistical tool for characterizing accuracy of medical diagnostic tests. Recently, several authors suggest approaches referred to as ROC regression models in order to evaluate effects of factors influencing accuracy of diagnostics. Oe, Ochi, & Goto (2016) have presented an inference process of a ROC regression model based on Generalized Additive Model with penalized splines via REstricted Maximum Likelihood method. This approach focuses on smoothing continuous-type covariates, but in addition to continuous covariates (e.g. age, weight, etc), discrete covariates (e.g. sex, smoking, etc) often significantly effect on accuracy of diagnostics. On this article, we proposed an extended method so that we could model discrete and continuous covariates simultaneously. We provided the formulation and the inference procedure, and evaluated inference performance of this method through several simulations. In the simulation result, we found that both effect of discrete and continuous covariates on the ROC was appropriately estimated. Furthermore, we applied our method to the neonatal hearing impairment screening data and evaluated how some covariates effect on the ROC with auditory brainstem response (ABR) as a diagnostic valuable for hearing impairment. From the analysis results, it was suggested that sex and Transient Evoked Otoacoustic Emission (TEOAE) influenced on the ROC with ABR.
Cognitive Diagnostic Models (CDMs) have been developed for a few decades to reveal students' knowledge status. The purpose of this paper is to provide a review of recent advances in CDMs, and to figure out the trend in its model development. In the paper, we first introduce the fundamental concept of CDMs, and then summarize broad range of modern CDMs by classifying them into compensatory, noncompensatory, and integrated models as well as other emerging approaches. Finally, we discuss the future orientation of CDM studies with focus on the empirical applications.
Experimental study with random allocation of treatments is regarded as the gold standard for estimation of causal parameters in statistical causal inference. However, when non-compliance among participants exists in the experiment it is likely to result in biased estimates of the parameters of interest. This paper discusses a sensitivity analysis in the estimation of causal parameters under apparent non-compliance. A motivating example is shown that concerns students' performance in university lectures. A procedure of sensitivity analysis for the estimates of causal parameters is presented that incorporates existence of apparent non-compliers, who are genuine compliers but act as non-compliers. An example of such procedure is shown for the case considered.
A maximum likelihood method for asymmetric multidimensional scaling, which was proposed by Saburi and Chino (Comput. Stat. Data Anal., 52:4673-4684, 2008), uses the additive error model in which the normally distributed error terms are added to the dissimilarities. In this study, we introduce in this method the multiplicative error model, where the log-normally distributed error terms are multiplied by the dissimilarities, and the corresponding representation of the dissimilarities. It was applied to the visibility data for the combinations of foreground and background colors, assuming the multiplicative as well as the additive error model. The optimal model was found with the multiplicative error model according to AIC.