Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2010. / ENGLISH ABSTRACT: The assessment of oral proficiency forms an important part of learning a second language. However, the manual assessment of oral proficiency is a labour intensive task requiring specific expertise. An automatic assessment system can reduce the cost and workload associated with this task. Although such systems are available, they are typically aimed towards assessing students of American or British English, making them poorly suited for speakers of South African English. Additionally, most research in this field is focussed on the assessment of foreign language students, while we investigate the assessment of second language students. These students can be expected to have more advanced skills in the target language than foreign language speakers. This thesis presents a number of scoring algorithms for the automatic assessment of oral proficiency. Experiments were conducted on a corpus of responses recorded during an automated oral test. These responses were rated for proficiency by a panel of raters based on five different rating scales. Automatic scoring algorithms were subsequently applied to the same utterances and their correlations with the human ratings determined. In contrast to the findings of other researchers, posterior likelihood scores were found to be ineffective as an indicator of proficiency for the corpus used in this study. Four different segmentation based algorithms were shown to be moderately correlated with human ratings, while scores based on the accuracy of a repeated prompt were found to be well correlated with human assessments. Finally, multiple linear regression was used to combine different scoring algorithms to predict human assessments. The correlations between human ratings and these score combinations ranged between 0.52 and 0.90. / AFRIKAANSE OPSOMMING: Die assessering van spraakvaardigheid is ’n belangrike komponent van die aanleer van ’n tweede taal. Die praktiese uitvoer van sodanige assessering is egter ’n arbeids-intensiewe taak wat spesifieke kundigheid vereis. Die gebruik van ’n outomatiese stelsel kan die koste en werkslading verbonde aan die assessering van ’n groot aantal studente drasties verminder. Hoewel sulke stelsels beskikbaar is, is dit tipies gemik op die assessering van studente wat Amerikaanse of Britse Engels wil aanleer, en is dus nie geskik vir sprekers van Suid Afrikaanse Engels nie. Verder is die meerderheid navorsing op hierdie gebied gefokus op die assessering van vreemde-taal sprekers, terwyl hierdie tesis die assessering van tweede-taal sprekers ondersoek. Dit is te wagte dat hierdie sprekers se spraakvaardighede meer gevorderd sal wees as di´e van vreemde-taal sprekers. Hierdie tesis behandel ’n aantal evaluasie-algoritmes vir die outomatiese assessering van spraakvaardighede. Die eksperimente is uitgevoer op ’n stel opnames van studente se antwoorde op ’n outomatiese spraaktoets. ’n Paneel van menslike beoordelaars het hierdie opnames geassesseer deur gebruik te maak van vyf verskillende punteskale. Dieselfde opnames is deur die outomatiese evaluasie-algoritmes verwerk, en die korrelasies tussen die beoordelaars se punte en die outomatiese evaluerings is bepaal. In kontras met die bestaande navorsing, is daar gevind dat posterieure waarskynlikheidsalgoritmes nie ’n goeie aanduiding van spraakvaardighede gee vir ons datastel nie. Vier algoritmes wat van segmentasies gebruik maak, is ook ondersoek. Die evaluerings van hierdie algoritmes het redelike korrelasie getoon met die punte wat deur die beoordelaars toegeken is. Voorts is drie algoritmes ondersoek wat daarop gemik is om die akkuraatheid van herhaalde sinne te bepaal. Die evaluerings van hierdie algoritmes het goed gekorreleer met die punte wat deur die beoordelaars toegeken is. Laastens is liniˆere regressie gebruik om verskillende outomatiese evaluerings te kombineer en sodoende beoordelaars se punte te voorspel. Die korrelasies tussen hierdie kombinasies en die punte wat deur beoordelaars toegeken is, het gewissel tussen 0.52 en 0.90.

Previous research has suggested that individuals differ in their accuracy of perceptions of the social environment, and some research suggests that powerful individuals in particular tend to be lazy, disinterested observers of the social world. A handful of field studies and lab experiments linking power with individuals’ perceptions of others’ social networks have generally supported this view. However, recent theory addressing the psychological consequences of power for the power-holder claim that in certain circumstances and for some kinds of social information, power is linked to increased accuracy of social information. This dissertation tests this idea by drawing on social network theory and the situated focus theory of power. I examine the relationship between individuals’ formal power and their perceptual accuracy of social network relationships. I propose that individuals’ perceptual accuracy is affected by 1) their formal power in the organization 2) the type of relationship being perceived (expressive/instrumental, positive/negative), and 3) the dependence relationship with the target of perception (whether the perceiver is dependent on the perceived to get their work done). Predictions were tested using cognitive social network data collected from a call center within a division of a large corporation in the US. Results showed that formal power was linked to increased accuracy for some relationship content (particularly negative expressive relationships), and managers tend to be more accurate when perceiving their own incoming relationships than non-managers.

<p>The risk assessment process and the subsequent risk management measures need tobe constantly evaluated, updated and improved. This thesis contributes to that workby considering, and suggesting improvements, regarding aspects like userfriendliness,transparency, accuracy, consistency, data reporting, data selection anddata evaluation.The first paper in this thesis reports from an empirical investigation of themotivations, intentions and expectations underlying the development andimplementation of a voluntary industry owned environmental classification systemfor pharmaceuticals. The results show that the purpose of the classification systemis to provide information, no other risk reduction measures are aimed for.The second paper reports from an evaluation of the accuracy and the consistency ofthe environmental risk assessments conducted within the classification system. Theresults show that the guideline recommendations were not followed in several casesand consequently alternative risk ratios could be determined for six of the 36pharmaceutical substances selected for evaluation in this study. When additionaldata from the open scientific literature was included the risk ratio was altered formore than one-third of the risk assessments. Seven of the 36 substances wereassessed and classified by more than one risk assessor. In two of the seven cases,different producers classified the same substance into different classificationcategories.The third paper addresses the question whether non-standard ecotoxicity data couldbe used systematically in environmental risk assessments of pharmaceuticals. Fourdifferent evaluation methods were used to evaluate nine non-standard studies. Theevaluation result from the different methods varied at surprisingly high rate and theevaluation of the non-standard data concluded that the reliability of the data wasgenerally low.</p> / QC 20100929

This study examines the messages and accuracy of the messages sent to students from the Arizona Instrument to Measure Standards (AIMS) math test regarding academic preparedness for postsecondary education. Previous studies investigating messages sent to students, derived from information such as educational standards, grades, course taking policies, test material, and college admissions requirements, focused on content alignment of secondary and postsecondary content standards. However, a dearth of research exists on messages from high school performance, as measured by exit exams and college performance measured by grades, major selection, or graduation. This study addresses the need to understand and interpret messages students receive based on academic performance. Specifically, this study aims to answer three questions. First, what is the relationship between AIMS math scores and college math performance defined by the University of Arizona math requirement and college graduation? Second, to what degree do AIMS math scores predict college math performance? Third, what is the average AIMS math score and performance level for students who choose certain majors? To answer the research questions posed in this study, various statistical analyses were employed. To answer the first question, a one-way ANOVA and logistic regression analyses were used. A linear regression analysis served to analyze the second and third questions. Results indicate that the messages sent to students regarding college readiness are, in fact, well aligned and clear and consistent. Also, there is evidence that the messages vary by gender and ethnicity.

本研究以I/B/E/S中2004至2005年所有美國公司為樣本，依財務分析師盈餘預測值與所有財務分析師盈餘預測平均值之差異程度，將財務分析師分類為大膽或膽怯的財務分析師，並依財務分析師盈餘預測發布之時點將財務分析師分類為領導型或從眾型財務分析師。針對財務分析師盈餘預測行為大膽及領導特性，探討時效性領導型財務分析師之盈餘預測是否會較大膽，並進一步研究大膽的財務分析師其盈餘預測準確度是否較高，以及時效性之領導型財務分析師其盈餘預測準確度是否較高。 研究結果發現時效性之領導型財務分析師其盈餘預測行為會較大膽，但大膽的財務分析師其盈餘預測準確度較低，且時效性之領導型財務分析師其盈餘預測準確度並未出現較高的現象。 / Security analysts can be characterized as bold or herding based on the absolute distance between their earnings forecasts and the consensus forecast. Security analysts can also be classified as lead or following based on the timeliness of their earnings forecasts. Based on I/B/E/S annual earnings forecasts of all American companies during the period of 2004-2005, this study addresses the association between bold and lead and the relation between bold forecast and forecast accuracy. In addition, the relation between lead forecast and forecast accuracy is investigated as well. It is shown that lead analysts are bolder than following analysts and boldness likelihood increases with the frequency of analysts’ forecast and declines with the analysts’ prior accuracy. Further, bold analysts’ earnings forecasts are less accurate than herding analysts’ and lead analysts’ earnings forecasts are less accurate than following analysts.

No description available.

Background: Insufficient exposure assessment is a suggested contributing factor to the current lack of clearly characterised relationships between occupational biomechanical risk factors and musculoskeletal disorders. Minimal attention has been paid to the potential bias of measurement tools from expected true values (i.e. accuracy) or between measurement tools, and empirical data on the magnitudes of variance contributed by methodological factors for measurement tool precision are lacking. Aim: The aim of this thesis was to quantify aspects of bias and precision in three commonly employed biomechanical risk factor assessment tools - inclinometry, observation, and electromyography (EMG) - and provide recommendations guiding their use. Methods: Upper arm elevation angles (UAEAs) were assessed using inclinometers (INC) and by computer-based posture-matching observation, and bias relative to true angles was calculated. Calibration models were developed for INC data, and their efficacy in correcting measurement bias was evaluated. The total variance of trapezius and erector spinae (ES) EMG recordings during cyclic occupational work was partitioned into biological and methodological sources, including the variance uniquely attributable to sub-maximal normalisation. Using algorithms to estimate the precision of a group mean, the efficacy of different trapezius EMG study designs was evaluated. Using precision criteria, the efficacy of different normalisation methods was assessed for ES EMG recordings. Results and Discussion: Inclinometer measured UAEAs were biased from true angles, with increasing bias at higher angles. In contrast, computer based posture-matching observations were not biased from true angles.  Calibration models proved effective at minimizing INC data bias. The dispersion of estimates between- and within- observers at any given set angle underlined the importance of repeated observations when estimating UAEAs.  For EMG, a unique but relatively small component of the total variance was attributable to the methodological process of normalisation. Performing three repeats of the trapezius EMG normalisation task proved optimal at minimizing variance for one-day EMG studies, while two repeats sufficed for multi-day EMG studies. A prone normalisation task proved superior for maximizing normalised lumbar ES EMG precision. Conclusion: Key aspects of measurement tool accuracy, bias between tools, and tool precision were quantified, and recommendations were made to guide future research study design.

Title: Influence of physical demands on shooting accuracy with hand-operated weapon. Objectives: Aim of this thesis is compare how physical load, present by Burpee motor abilities test, aplicate on selected group of soldiers influence their shooting accuracy from pistol in stand up position with bouth hands grip at standing target. Aggregate of twenty test subjects is comosed of policemen from Department of protect and escort Military police Prague and students from VO at FTVS UK in Prague. Methods: For statistical data processing we used the method of comparison and comparison of averages. To determine statistical significance, we used two- tailed, paired t - test. For the evaluation targets, we used numerical method for detecting mean point of impact. Results: The measured data were processed statistically, it was found that physical activity has statistical effect on the shooting accuracy, while standing and with both hands grip, from pistole in the first group (MP). The second measurement work out similary, which mean physical exercise has statistical effect on the shooting accuracy, while standing and with both hands grip, from pistole. Furthermore, we found that the greatest influence on shooting accuracy after physical exertion realized by Burpee test, have the individual level of fitness...

Master of Science / Department of Biological and Agricultural Engineering / Stacy L. Hutchinson / Military training lands are a vital resource for national security and provide crucial habitat for a number of threatened and endangered species. Military land managers must manage the land in accordance with federal environmental policy and regulation, while simultaneously providing the lands needed for training military forces. Off road maneuver training can cause significant environmental damage including removal of vegetation, compaction of soils, increased erosion, loss of habitat, and degradation of the landscape to a point of not being useful for continued military training. Various techniques have been developed to help the military land managers determine a sustainable training level for the landscape. Many of these techniques have limitations in the spatial resolution of data collected and the ability to provide timely and accurate assessments of training disturbance. Advancements in GPS and GIS technology over the past two decades have shown the potential to fill this knowledge gap. In this study low cost civilian off the shelf (COTS) GPS devices were accuracy tested to determine their capability to provide reliable and accurate military vehicle locations during training (1.93 m CEP, 4.625m 2dRMS). The GPS data collected from COTS devices on three battalion training exercises at Fort Riley, KS were processed in a GIS and statistically analyzed to compare and contrast several off road maneuver metrics (speed, turning radius, distance traveled) by vehicle type tracked, and by platoon in order to determine if units or vehicle types could reliably explain the variation in these metrics. Lastly, a method of mapping the relative environmental disturbance was developed and mapped for the same data sets. Wheel sinkage was used as a measure of disturbance, it was calculated at each GPS point based on vehicle type and soil conditions then mapped in using a fishnet grid for Fort Riley, Kansas.

A presente dissertação teve como finalidade avaliar a precisão e a discrepância das coordenadas (acurácia) obtidas por meio das observações GPS, no modo estático rápido, medidas com receptores de uma freqüência, empregando-se a técnica de suavização do código C/A pela fase da portadora para diferentes comprimentos de linha base (20, 50, 75, 100, 150, 200 e 300 Km), a partir da estação de monitoramento contínuo da EESC/USP. Essa técnica obteve boa repercussão no Brasil, a partir do momento em que foi permitida sua utilização para fins de georreferenciamento de imóveis rurais, conforme a norma técnica do INCRA, que permite a utilização da suavização do código pela fase da portadora, desde que esta apresente acurácia com valores iguais ou inferiores a 50 cm. Atualmente, trabalhos com tal técnica estão sendo desenvolvidos e aceitos para compor o banco de dados do cadastro rural nacional. Pelos resultados obtidos nesta dissertação, foi possível avaliar que a aplicação dessa técnica não atende à exigência requerida em 100% das medições e, ao mesmo tempo, verificar a diferença entre a precisão do pós-processamento, indicada pelos softwares comerciais e a acurácia dos pontos medidos comparados com os pontos de controle estabelecidos para esse fim. / The present dissertation has as objective to evaluate precision and discrepancies of coordinates (accuracy) obtained through GPS observations, in a rapid static method, measured by single frequency, which applies carrier-smoothed-code for different baseline length (20, 50, 75, 100, 150, 200 and 300 km), since continuous monitoring station from EESC/USP. This technique obtained a good repercussion in Brazil, since when its use was legitimate in rural property georeferencing, according to INCRA\'s norma técnica (technical rule), which allows the use of carrier-smoothed-code in view of the fact that presents accuracy with equal or inferior values to 50 cm. Currently, works using such technique are developed and accepted in order to compound database of national rural survey in Brazil. Based on the obtained results it was possible to evaluate that the technique does not guaranteed to requested requirement in 100% of measurements, and at the same time to verify the difference between post-processing precision, indicated by commercial software, as well as measured point accuracy compared with points established for this purposed.