No description available.

No description available.

This dissertation illustrates how to detect the rater centrality effect in a simulation study that approximates data collected in large scale performance assessment settings. It addresses three research questions that: (1) which of several centrality-detection indices are most sensitive to the difference between effect raters and non-effect raters; (2) how accurate (and inaccurate), in terms of Type I error rate and statistical power, each centrality-detection index is in flagging effect raters; and (3) how the features of the data collection design (i.e., the independent variables including the level of centrality strength, the double-scoring rate, and the number of raters and ratees) influence the accuracy of rater classifications by these centrality-detection indices. The results reveal that the measure-residual correlation, the expected-residual correlation, and the standardized deviation of assigned scores perform better than the point-measure correlation. The mean-square fit statistics, traditionally viewed as potential indicators of rater centrality, perform poorly in terms of differentiating central raters from normal raters. Along with the rater slope index, the mean-square fit statistics did not appear to be sensitive to the rater centrality effect. All of these indices provided reasonable protection against Type I errors when all responses were double scored, and that higher statistical power was achieved when responses were 100% double scored in comparison to only 10% being double scored. With a consideration on balancing both Type I error and statistical power, I recommend the measure-residual correlation and the expected-residual correlation for detecting the centrality effect. I suggest using the point-measure correlation only when responses are 100% double scored. The four parameters evaluated in the experimental simulations had different impact on the accuracy of rater classification. The results show that improving the classification accuracy for non-effect raters may come at a cost of reducing the classification accuracy for effect raters. Some simple guidelines for the expected impact of classification accuracy when a higher-order interaction exists summarized from the analyses offer a glimpse of the "pros" and "cons" in adjusting the magnitude of the parameters when we evaluate the impact of the four experimental parameters on the outcomes of rater classification. / Ph. D.

當研究者想要了解態度量表中不同組間之態度分數是否有所差異時，一個常見的分析方法為變異數分析。然而，變異數分析需要建立在資料服從常態分配之假設上，態度量表之資料類型卻很明顯地不符合此一假設。而非針對連續型資料所推導出來的 統計量，應該是較適合處理序列或是等距尺度等非常態資料之檢定方法。本研究主要之目的即為探討利用 統計量以及利用變異數分析兩者所作出之檢定結果差異為何。過去相關研究皆假設態度量表背後存在一連續潛在變數，本研究則直接由間斷型分配出發。在公式推導上，我們發現 統計量與變異數分析中之 統計量存在一對一對應之關係。雖然兩統計量近似之分配不同，但兩統計量所對應之p值卻始終非常接近。若以0.05為顯著水準， 統計量與 統計量之檢定結果幾乎完全相同。當需要檢定不同組間在多題上之看法是否具有差異時，我們比較了將屬於同一主題之各題分數加總，然後依照單變量變異數分析之方法進行檢定，以及多變量變異數分析法、羅吉斯迴歸分析法等三種方法。根據我們的模擬結果，若各組在各題之態度皆很類似，則利用ANOVA進行分析可以得到較低的型一誤差；若各組在各題之態度不太一致，且有左右偏分配互相抵銷的情形，則利用MANOVA或是羅吉斯迴歸分析法才能夠維持住很高的檢定力。 / In social science literature, we frequently found that ANOVA techniques were utilized to analyze Likert-type response data. However, one of the three basic assumptions behind ANOVA is that response variable is normally distributed, and Likert-type data apparently do not share this property. In this study, we compare the performance between statistic associated with ANOVA with Mantel- Haenszel statistic, a statistic aimed at handling categorical data. We found that statistic and statistic have one-to-one relationship. Although these two statistics can be approximated by distribution and Chi-square distribution respectively, their p values are quite close to each other. At the significant level of 0.05, and statistics almost have the same testing results. In addition to analyzing a single Likert-type response question, we would also like to analyze a set of Likert-type response questions that probably represent a specific concept. We propose two alternatives here. The first one is MANOVA, and the second one is logistic regression analysis. According to the simulation results, using the ANOVA approach is slightly better in terms of the type I error rate if the responses have similar structures among questions. On the other hand, using MANOVA or logistic regression analysis would maintain higher power whenever the responses have different structures among questions.

The purpose of the study was to investigate how the online, computer-based learner's personal learning profile (Preference), the content delivery method supplemented with visual content based on Neil Fleming's VARK (visual, aural, read/write, kinesthetic) model (Content), and the interaction of Preference and Content, influenced learner performance (Performance) and/or learner self-reported satisfaction (Satisfaction). Participants were drawn from a population of undergraduates enrolled in a large public southwestern research university during the fall 2015 semester. The 165 student participants (13.79% completion rate) were comprised of 52 (31.5%) females and 113 (68.5%) males age 18-58+ years with 126 (76.4%) age 18-24 years. For race/ethnicity, participants self-identified as 1 (0.66%) American Indian/Alaska Native, 21 (12.7%) Asian/Pacific Islander, 27 (16.4%) Black, non-Hispanic, 28 (17%) Hispanic, 78 (47.3%) White, non-Hispanic, 10 (6.1%) other. Reported socioeconomic status was 22 (13.3%) withheld, 53 (32.1%) did not know, 45 (27.3%) low, 13 (7.9%) moderately low, 16 (9.7%) middle, 8 (4.8%) upper middle, and 8 (4.8%) upper. This causal-comparative and quasi-experimental, mixed-method, longitudinal study used researcher-developed web-based modules to measure Performance and Satisfaction, and used the criterion p < .05 for statistical significance. A two-way, 4 x 3 repeated measures (Time) analysis of variance (RM-ANOVA) using Preference and Content was statistically significant on each Performance measure over Time, and at two measures on Satisfaction over Time. The RM-ANOVA was statistically significant on between-subjects main effect Performance for read/write modality Content compared to aural and kinesthetic Content. There were no statistically significant main effects observed for Satisfaction. A Pearson r correlation analysis showed that participants that were older, married, and of higher socioeconomic status performed better. The correlation analysis also showed that participants who performed better reported greater likelihood to take online courses in the future, higher motivation, sufficient time and support for studies, and sufficient funding for and access to the Internet. The study results suggested that regardless of Preference, using read/write modality Content based on the VARK model while maintaining the verbal language can yield better Performance outcomes. The study results also suggested that while maintaining the verbal language, Preference, and Content based on the VARK model do not distinguish learner Satisfaction outcomes. However, because Satisfaction has been shown to impact Performance, efficacy, and retention, it matters to educational institutions. Future research should consider more granular models and factorial research methods, because models that utilize a single representative construct score can mask effects when analyzing Performance and Satisfaction.

This study attempted to identify ways to improve voluntary compliance and minimize taxpayer burden, but also tries to understand the behavior of taxpayers’ compliance given the compliance regimen. Most explicitly, it attempted to identify ways to improve payment compliance using regimens already utilized in other parts of the tax code. The research question was whether different tax regimes, such as safe harbor, can change the behavior of employment tax payment for small business or self-employed taxpayers. The idea was to determine if a safe harbor provision can reduce the proclivity of authorized individuals to implement a payroll tax dilemma strategy and whether or regimen can reduce payment noncompliance in time of economic distress. To answer the research question, an online experiment was employed. The experimental design was an impact study. The population of interest in this study was all authorized individuals from small and self-employed firms. The sample size totaled 205, and it was based on the a-priori sample size calculation. Analysis of variance (ANOVA) was chosen as the data analysis technique, but other nonparametric test and logistic regression models were used to further analyze the data. This study showed that for subjects who did not subscribed to safe harbor provision but experienced an increased probability of apprehension increased their payment compliance. Also, the availability of a safe harbor provision lead to a large numbers to a safe harbor provision subscription in order to avoid enforcement. This study was able to show that individuals were willing to improve their payment compliance rate when enforcement was increased. The General Deterrence Theory explains that increased deterrence will lead to higher compliance. The study showed a 10 percent improvement in payment compliance when safe harbor was implemented. The results from this study also suggest that provisions such as a safe harbor can be a method of reducing filing costs and audit costs and ultimately taxpayer burden. On the other hand, the results of this study were inconclusive in determining if such provisions can improve payment compliance. Nevertheless, the outcome of this study can improve timing and accuracy of employment taxes payments and it may improve the accuracy of employment tax payment.

På grund av den globalt snabba tillväxten av mänskliga samhällen har mängden artificiellt ljus ökat. Artificiella ljuset hotar fladdermössens överlevnad på grund av de störningar som ljuset orsakar i fladdermössens födosökande, parning, habitat samt deras cirkadianska system. Syftet med denna studie är att se om det finns ett samband mellan förekomsten av fladdermöss och ljusföroreningar i Stockholms län samt om förekomsten av ljusopportunistiska respektive ljuskänsliga fladdermöss skiljer sig vid olika grader av ljusföroreningar. I studien delades Stockholm i tre stycken zoner: zon 1 har mest ljusföroreningar, zon 2 har mindre, och zon 3 har minst. Styrkan på ljusföroreningar mättes med Sky Quality Meter version L vid 15 stycken fladdermusinventeringar i vardera zonen. Fladdermusarterna delades in i två grupper, ljusopportunistiska och ljuskänsliga. Resultatet av studien visar ingen signifikant interaktion mellan fladdermusgrupp och zonindelningens påverkan på fladdermusandel. Ljusföroreningarna indikerar ingen signifikant påverkan på fladdermössantal. Variablerna ljusförorening, avstånd till väg och järnväg samt molnighet har ingen signifikant påverkan på fladdermössantalet.

The physical properties of polymer quench bath directly affect the cooling rate of a quenched part. These properties include the type of quenchant, its temperature, concentration, and agitation level. These parameters must be controlled to optimize the quenching process in terms of alloy microstructure, properties and performance. Statistically designed experiments have been performed to investigate the effects of the process parameters (i.e. polymer concentration and agitation) on the heat transfer behavior of cast aluminum alloy A356 in aqueous solution of Aqua-Quench 260 using the CHTE quenching-agitation system. The experiments were designed using Taguchi technique and the experimental results were analyzed with Analysis of Variance (ANOVA) based on the average cooling rate. It is found that average cooling rate dramatically decreases with the increase in polymer concentration. Agitation only enhances the average cooling rate at low and medium concentration levels. From ANOVA analysis, the process parameter that affects the variation of average cooling rate most is the polymer concentration, its percentage contribution is 97%. The effects from agitation and the interaction between polymer concentration and tank agitation are insignificant. The mechanical properties of age-hardenable Al-Si-Mg alloys depend on the rate at which the alloy is cooled after the solutionizing heat treatment. A model based on the transformation kinetics is needed for the design engineer to quantify the effects of quenching rates on the as-aged properties. Quench Factor analysis was developed by Staley to describe the relationship between the cooling rate and the mechanical properties of an age-hardenable alloy. This method has been previously used to successfully predict yield strength, hardness of wrought aluminum alloys. However, the Quench Factor data for aluminum castings are still rare in the literature. In this study, the Jominy End Quench method was used to experimentally collect the time-temperature and hardness data as the inputs for Quench Factor modeling. Multiple linear regression analysis was performed on the experimental data to estimate the kinetic parameters during quenching. Time-Temperature-Property curves of cast aluminum alloy A356 were generated using the estimated kinetic parameters. Experimental verification was performed on a L5 lost foam cast engine head. The predicted hardness agreed well with that experimentally measured.

This study investigates how outliers of different magnitude impact the robustness of the twosample t-test. A simulation study approach is used to analyze the behavior of type I error rates when outliers are added to generated data. Outliers may distort parameter estimates such as the mean and variance and cause misleading test results. Previous research has shown that Welch’s ttest performs better than the traditional Student’s t-test when group variances are unequal. Therefore these two alternative statistics are compared in terms of type I error rates when outliers are added to the samples. The results show that control of type I error rates can be maintained in the presence of a single outlier. Depending on the magnitude of the outlier and the sample size, there are scenarios where the t-test is robust. However, the sensitivity of the t-test is illustrated by deteriorating type I error rates when more than one outlier are included. The comparison between Welch’s t-test and Student’s t-test shows that the former is marginally more robust against outlier influence.

雖然允許變異數－共變數矩陣異質之LISREL-MANOVA模式能克服MANOVA在理論上的的限制（變異數－共變數矩陣同質性的假設），而帶有潛在變項的LISREL-MANOVA能解決MANOVA分析目的上的限制（不能整體考驗多組在潛在變項平均數向量上的差異），但仍有一些問題尚未解決與釐清。其一是MANOVA後續分析的問題，其二是帶有潛在變項之LISRES-MANOVA的統計特性以及切割現象所造成的影響。據此，本論文的主要研究目的有二：（一）發展允許變異數－共變數矩陣異質之LISREL-MANOVA的部份考驗模式以解決MANOVA後續分析中同質性假設的限制，並以模擬分析的方式驗證之；（二）利用模擬分析的方式來探討帶有潛在變項之LISREL-MANOVA的統計特性。 　　依據上述研究目的，本論文具體的研究問題有二：（一）在變異數－變異數－共變數數矩陣同質或變異數－變異數－共變數數矩陣異質的狀態下，允許變異數－共變數異質之LISREL-MANOVA的部份考驗和傳統MANOVA的後續分析（單變量F考驗、區別函數標準化加權和典型變量相關）何者較為優異？（二）是否有切割現象、切割的程度、樣本大小以及估計方法等因子對帶有潛在變項之LISREL-MANOVA的估計有何影響？ 　　針對上述研究問題，本論文主要的研究結果有二：（一）在變異數－共變數同質的狀態下，允許變異數－共變數異質之LISREL-MANOVA的估計結果與單變量F考驗、典型變量相關一致，但較區別函數標準化加權顯著的優異；在變異數－共變數異質的狀態下，允許變異數－共變數異異質之LISREL-MANOVA的估計結果較單變量F考驗、區別函數標準化加權和典型變量相關優異。（二）就帶有潛在變項LISREL-MANOVA而言：1.切割現象會使參數估計值較不精確因而使整體配適考驗容易達顯著。2.切割的比例越大會使整體配適考驗達顯著的機率急遽的上升。3.在切割比例固定下，樣本越大越容易使整體配適考驗達顯著，且會增加WLS估計法犯型Ⅰ錯誤的可能性及GLS和ML估計法的統計考驗力。以及4.在GLS、ML和WLS三種估計法中，WLS較不容易使整配適考驗達顯著，但較容易犯型Ⅰ錯誤；GLS和ML較容易使整體配適考驗達顯著且容易犯型Ⅱ錯誤。 　　最後，本論文根據上述研究結果做成結論，並對應用上及未來研究方向上提出建議。