141 |
The Connection Between External Environment and Internal Strategy : a case study of Scandinavian Airlines SystemChen, Ziyuan, Liao, Zijun January 2009 (has links)
A variety of factors are the reasons for adjusting or changing company’s strategy, such as the change of customer demand, company’s internal financial factors, the influence of external environment and so on. This research is to find out the link between the change of external environment and the adjustment of internal strategy. Choose the suspension of one flight in SAS as the case to state this point. Show the strategy change of SAS when they were facing the serious impact of financial crisis. SAS as the biggest airline in North Europe it could be a typical firm to investigate this kind of situation. A deep-interview and reading the published reports of SAS are the main data collection approached. Some resources were from the internet because it is real-time news and the reports from the company are published on their website. The interviewee is from the top of the company who is familiar this area and has full experience in international business. The study used lots of theories from different books and journals to integrate the information that we collected to analysis and achieve our final conclusion. After analyzed the study it found the SAS used the retrenchment strategy as their new international strategy to reverse the negative situation. The financial crisis affected the customer demand badly not just in China also around the world and forced them to change the strategy. The report ultimate believed there were still a lot of other factors, beside the financial crisis, caused the adjustment of strategy in SAS.
|
142 |
Oljeprisförändring och dess inverkan på flygbranschenIbrahim, Akil, Nazem-Z, Mohammad January 2006 (has links)
Sammanfattning Bakgrund Råoljor utgör den råvarumässiga grunden för cirka 40 % av den kommersiella energi- marknaden i världen. Därför oljemarknaden har en betydande ekonomisk och politisk roll i världen och betraktas priset på råolja idag som en viktig faktor för utveckling av världsekonomin. På senare tid har alla oroligheter i Mellanöstern, bland annat kriget i Irak och problematiken kring Irans kärnkraftverksamheter samt en ovanligt aktiv inledning på orkansäsongen i sydöstra Förenta staterna under 2005 drivit upp ett redan högt råoljepris. Enligt insatta aktörer på marknaden är flygbranschen en av de beroende branscherna till oljesektorn. Den är beroende av oljeproduktion, oljepris och oljetillgänglighet. Det påstås att oljeprishöjningen kan leda till högre kostnadsposter och följaktligen negativ inverkan på flygbolagens resultat. En ansenlig del av flygbranschens kostnader utgörs av bränslekostnader och dessa uppgår för närvarande till mellan 10 och 25% för de flesta flygbolag. Det gör att varje prisförändring i oljebranschen kan påverka flygbranschens lönsamhet. Att se till vilken grad dessa påståenden motsvarar det verkliga förhållandet när det gäller oljepriset och dess påverkan på flygbolagen, kommer att undersökas i denna uppsats. Syfte Syftet är att undersöka och analysera hur prisförändringar i oljebranschen påverkar Scandinavian Airlines SAS och Ryanairs resultat. Metod Undersökningen kommer att göras i form av en fallstudie och omfattar ovannämnda två flygbolag å ena sidan och oljepriset å den andra. Undersökningen genomförs med olika metoder såsom analys av texter och intervjuer. Resultat Enligt undersökningen är korrelationen mellan oljesektorn och flygbranschen negativ när det gäller bolagens bränslekostnader, men inte beträffande bolagens resultat. Det vill säga: när oljepriset stiger, påverkas flygbolagen SAS och Ryanair i form av ökade bränslekostnads- poster. Men genom effektivisering, det slutliga resultatet för både bolagen är positivt. / Background Crude oil constitutes 40% of basic raw material of commercial energy market in the world. Therefore the oil market has a meaningful economical and political role in the world. This makes the price on the oil market to become an important factor for the development of the world's economy. Lately the political instability in the Middle East, especially the war in Iraq, the complexes of the problems in Iran's nuclear weapon program and the unusual active beginning of the hurricane season in the south-eastern areas of the United States of America in 2005 has driven up already high prices of the raw oil to even higher levels. According to market analysers the flight industry is one of those industries that depend on the oil section. It is addicted to the oil production, prices of oil and oil availability. According to some people the increase of oil prices can bring up high expense and therefore a negative influences on the results of the flight companies. One considerable expense of the flight industries is fuel expenditure. At the moment the cost for fuel expenditure for many flight companies is between 10 and 25%. These create any price exchange of oil industry to affect the profitability of the flight industry. The statement that have been mentioned above and how those are connected to the reality when it comes to oil prices and its effectiveness on the flight companies will be the basic of this study. Purpose The purpose of this study is to study and analyse price exchanges in the raw oil industry affects Scandinavian Airlines (SAS) and Ryanair's results. Method This study will conduct with a case study and will include the two above named flight companies. The study will be carrying out with different methods as analyse of text and interviews. Results The result of these study saws that the relation between the oil section and flight industry is negative when it comes to flight companies fuel cost, but the result does not saw any negative result on flight companies results. This means that any increase of oil prices will affect SAS and Ryanair fuel expenditures but the final result for these two flight companies through efficiency is positive.
|
143 |
An AHP framework for balancing efficiency and equity in the United States liver transplantation systemVeerachandran, Vijayachandran M 01 January 2006 (has links)
ABSRACT: Liver transplantation and allocation has been a controversial issue in the United States for decades. One of the main concerns in the allocation system is the trade-off between the two main objectives, efficiency and equity. Unfortunately, it is difficult to reach consensus on how to develop allocation policies that aim at balancing efficiency and equity, among transplantation policy makers, administrators, transplant surgeons and transplant candidates.Our research identifies and classifies the outcomes of liver allocation into two major categories, efficiency and equity, that are, often times, conflicting. Previous researchers did not consider how to balance outcomes in these two categories. Our research uses Analytic Hierarchy Process, a Multi-Criteria Decision Analysis methodology, to build a framework that quantifies the decision-making process and help decision makers to reach a valid consensus in terms of balancing these outcomes. Latest available patient registration and follow-up data are used in data analysis. Results from this analysis serve as inputs for the simulation model that is capable of evaluating alternative hypothetical policies.This research addresses the deficiencies of the current liver transplantation policy and is intended to refine the policy that will result in a more balanced allocation system with respect to efficiency and equity. Our proposed methodology can be applied to incorporate further changes in policy selection and refinement.
|
144 |
Development of a new drawing system for STSHåkonsen, Christian, Berkelund, Mikael January 2008 (has links)
An engineering firm which handles and constructs drawings needs well defined routines and structures which should be homogeneous through all the different departments. A common drawing system results in better quality and cooperation between the departments. SAS Technical Services (STS) did not have a common drawing system which had led to development of different routines in the different regions and departments. Requested was development of new routines regarding engineering drawings, such as drawing numbering structure, revision and subscription routines, which standards to adhere to, custom made drawing templates and management of the drawings with belonging documents. Each requested task was broken into minor tasks and analyzed. Solutions by different leading engineering companies were used for comparison and ideas. All the tasks were collected and organized in one single document which is the result of the thesis; a drawing instruction. The drawing instruction will after a learning phase ease the work for the STS engineers as all necessary information can be found in one single place. Also, work with contractors will be time-saving as the instruction can be handed out for guidance. / En ingenjörsfirma som hanterar och skapar mängder med ritningar behöver väldefinierade rutiner och strukturer som är homogena genom hela bolaget. Ett gemensamt ritningssystem resulterar i bättre kvalitet och bättre samarbete mellan de olika avdelningarna. SAS Technical Services (STS) hade inte ett gemensamt ritningssystem vilket har lett till att de olika avdelningarna har utvecklat olika rutiner rörande ritningar. Önskemålet var utvecklande av nya rutiner rörande ingenjörsritningar såsom ritningsnumreringsstruktur, revisions- och prenumerationsrutiner, vilka standarder som skall följas, egna ritningsmallar och skötsel av ritningar med tillhörande dokument. Varje önskad uppgift blev uppdelad i mindre uppgifter och analyserad. Lösningar från olika ledande ingenjörsbolag användes för jämförelse och idéer. Alla uppgifterna samlades och organiserades i ett och samma dokument som är resultatet av detta arbete; en ritningsinstruktion. Ritningsinstruktionen kommer efter en inlärningsfas underlätta arbetet för ingenjörerna på STS eftersom all nödvändig information nu finns på ett och samma ställe. Arbeten med entreprenörer vill också bli underlättande eftersom instruktionen kan delas ut för vägledning.
|
145 |
Defining the nature and outcomes of Australian professional supervision : applying Holloway's systems approachJohnston, Karla Gai January 2006 (has links)
The goal of this thesis was to define the nature and outcomes of Australian professional supervision by applying Holloway's (1995) Systems Approach to Supervision (SAS) across professional groups. Many Australian professionals such as psychologists, counsellors, and accountants are required to participate in some form of supervision before being granted permission, via registration, to practice independently within their respective fields. This is the first study of its kind to investigate the supervision experience of a range of professional groups within Australia. The SAS model (Holloway, 1995) provided a well-researched theoretical and practical framework with the potential to be applied across professional groups for the purposes of developing, evaluating and enhancing supervisor and supervisee practice. Based on the model, Holloway proposed a number of teaching tasks and functions in a matrix to explain the process of supervision and to assist in the professional and personal development of supervisees. To date, most of the literature on professional supervision has failed to provide a theoretical framework from which results could be meaningfully interpreted. This thesis outlines a program of research which used the SAS model as a theoretical basis for understanding and evaluating the experience of supervision amongst a range of Australian professional and its relationship to effective professional practice. Four studies were conducted in accordance with Mackenzie and House's (1979) Model of Scientific Inquiry. The first study (Study One) was exploratory in nature, and aimed to define the term "supervision" and the anticipated outcomes of supervision activities. The second study (Study Two) was also exploratory in nature, and aimed to evaluate the modes of supervision delivery as well as to collect information regarding the key tasks and functions utilised in professional supervision. The third study (Study Three) was empirical in nature, and investigated the supervision experience of psychologists engaged in supervision in accordance with the SAS model (Holloway, 1995). It longitudinally tracked their performance over a 12-month period as evaluated by both their supervisors and work managers. The fourth study (Study Four) was confirmatory in nature, and was the same as the third study but comprised a different sample of business and accounting (who were not Certified Public Accountants) graduates. The four studies consisted of four separate samples surveyed with self-report measures developed from the SAS model (Holloway, 1995) by the researcher. Study One comprised a sample of 210 supervisor-supervisee dyads. Professional groups in this sample were psychologists, counsellors, nurses, occupational therapists, financial advisors, business consultants, and accountants (without CPA) all of whom were participating in a supervision process. Study Two comprised a total of 200 supervisees broken down into four groups of 50. The professional types included in this sample were psychologists, counsellors, nurses and business consultants and accountants (without a CPA). Study Three comprised 513 supervisees who were participating in supervision as part of the criteria to become fully registered psychologists. Study Four included 480 business consultants and accountants (without CPA) who were in the early years of their career and were participating in supervision as part of their professional development. The central aims of the supervision experience, according to supervisees and supervisors, were to develop skills in counselling, case experience, professional experience, emotional awareness, the ability to self-evaluate and network. The six reported outcomes of supervision were that professional supervision enhanced supervisees' ability to self-evaluate, gain academic knowledge, become emotionally aware, develop profession networks, develop both professional and work skills and to build on relationship skills. The findings supported and extended the SAS model (Holloway, 1995) by adding the tasks of academic knowledge and networking. The definition of supervision found in this research program also supported and built on the definitions already provided in the literature. There was considerable support for the SAS model's matrix in that particular supervision functions employed to teach certain supervision tasks were more effective than others. For example, on the one hand, to teach a supervisee the skill of emotional awareness, a supervisor is best advised to take a supportive/sharing approach. On the other hand, it was shown that the teaching strategy of monitoring/evaluating was not found to be conducive to teaching case conceptualisation skills. The findings also suggested two enhancements to Holloway's original conceptualisation of the SAS model of supervision delivery mode and supervisor allocation. Furthermore, the findings confirmed that the SAS model can be applied to teach and objectively evaluate supervision success by supervisors and managers across professional groups. There were six major contributions of this research program to the field of professional supervision: First, there was the application and validation of a theoretical model, Holloway's (1995) Systems Approach to Supervision, to the supervision experience. Second, there was the development and application of scales to measure supervision performance and satisfaction reliably and with demonstrated construct validity. Third, the methodology which included the collection of both qualitative and quantitative responses from supervisees, supervisors, and managers provided a multi-method approach to understanding professional supervision across professional groups. Fourth, a uniform definition of supervision was identified across a range of professional groups. Fifth, Holloway's supervision teaching matrix was empirically supported and the findings recommend it as a mechanism for developing, evaluating and enhancing supervisor and supervisee practice across a range of professional types. The SAS model was found to be relevant to health-related professional groups but also supported in business-related professional groups. Finally, the research recommended some modifications to the SAS model to incorporate factors such as supervision delivery mode. These additions and the results of the longitudinal research suggested that supervision effectiveness was a predictor of on the job performance ratings by managers. Based on a scan of the literature to date, this research program outlines the first longitudinal empirical study of the relationship between supervision effectiveness and on-the-job performance using a range of professional groups. Limitations and future directions were discussed.
|
146 |
Análise de experimentos industriais com respostas categóricas ordenadas : método de Taguchi e modelo de McCullaghVigo, Álvaro January 1994 (has links)
Nas diversas áreas da ciência e na Tecnologia da Qualidade são realizados experimentos para estudar a influência de diversos fatores sobre características (tais como "severidade do desgaste" ou "qualidade de uma solda"), que são registradas como variáveis categóricas ordenadas. Nessa dissertação são estudados aspectos da análise de experimentos com resposta ordinal, como ferramenta da Tecnologia da Qualidade. Os métodos tradicionais de análise estatística (teste Qui-quadrado e testes não-paramétricos, por exemplo) não são eficientes nessa situação, sendo necessários métodos mais elaborados para extrair e interpretar as informações geradas nesses experimentos. Diante disso, foram estudados em detalhe dois métodos alternativos: o primeiro é a técnica da Análise de Acumulação (AA) introduzida por Taguchi (I 987) e, o outro método, baseia-se no modelo de Odds Proporcionais proposto por McCullagh (1980), através do qual é possível estimar e interpretar os efeitos dos fatores. Uma característica especial dessa resposta ordinal é a possibilidade de estar associada a uma variável contínua latente. Assim, os dados observados podem ser vistos como uma categorização dessa variável contínua, não diretamente observável. Essas técnicas de análise são ilustradas para o caso unifatorial e multifatorial, mediante dois conjuntos de dados apresentados na literatura: um estudo observacional e outro experimental. O ajuste dos modelos foi realizado através do procedimento LOGISTIC do pacote estatístico SAS e pelo software ST ATA. Com base nos resultados obtidos, verifica-se a necessidade de modelos paramétricas para estimar magnitude e direção dos efeitos que os fatores provocam na característica de qualidade, uma vez que os métodos tradicionais e a AA não extraem informação suficiente dos dados. Nesse sentido, o modelo de Odds Proporcionais parece um método simples, adequado e eficiente para a análise de dados com resposta ordinal. / In several areas o f Science and Quality o f Technology, experiments are conducted to study the influence of severa! factors on characteristics (as "worn-out severity, or "weld quality"), that are recorded as ordered categorical variables. In this thesis, some aspects o f the analysis o f experiments with ordinal response are studied as a tool of Quality of Technology. The traditional methods of statistical analysis (Chísquare test and non-parametric tests, for example) are not efficient in thís situation, and more sophisticated methods are needed to extract and to interpret the information produced in these experiments. So, two alternative methods are studied in detail: the first is the Accumulation Analysis (AA) technique introduced by Taguchí (1987). The other method is based on the Proportional Odds model proposed by McCullagh (1980), whích allows to estimate and to interpret the effects of the factors. A special characteristics of this response is that it might be associated to a continuous Iatent variable. Thus, the observed data may be seen as a categorization o f this continuous variable not directly observable. These techniques of analysis are illustrated for the single-factor and the multifactor cases, by using two data set presented in the literature: an observational study and an experimental study. The models were adjusted by the procedure LOGISTIC o f the software SAS and by the software ST ATA. The results obtained point to the need of parametric models to estimate the magnitude and direction of the effects caused by the factors in the characteristic of quality, since the traditional methods and the AA do not extract enough information of the data. In thís sense, the Proportional Odds model seems a simple, adequate and efficient method for analyzing data with ordinal response.
|
147 |
A utilização de métodos de regressão para respostas ordinais no estudo de coorte das crianças nascidas em 1993 em Pelotas, RS, BrasilNunes, Luciana Neves January 1999 (has links)
Em estudos epidemiológicos. Os desfechos são freqüentemente medidos em escalas ordinais, entretanto são analisados por técnicas tradicionais. como a reg ressão logística que requer desfecho binário. Consequentemente. informações importantes podem ser perdidas devido a arbitrariedade na escolha do ponto de corte. Modelos de regressão para respostas ordinais têm sido desenvolvidos e têm vantagens de reter a ordinalidade dos dados e fornecer estimativas interpretáveis para os epidemiologistas. Esse artigo descreve o modelo de odds proporcionais proposto por McGullagh e o aplica aos dados da coorte de cnanças nascidas em Pelotas em 1993. usando o pacote estatístico SAS. O objetivo desse artigo foi comparar os resultados do modelo de odds proporcionais e da regressão logística tradicional usando diferentes pontos de corte. O desfecho foi déficit de estatura para idade expressa em escore-z. com três níveis ordinais. Concluiu-se que o modelo de odds proporcionais produz uma estimativa interpretável. similar a uma razão de odds. que sumariza o efeito sobre todos pontos de corte. Na presente análise. as estimativas pontuais resultantes da regressão ordinal foram similares ás obtidas através da regressão logística tradicional usando o ponto de corte do escore-z em -1. Entretanto, a regressão ordinal produziu intervalos de confiança menores (e, portanto, aumento do poder estatístico), evitando a arbitrariedade de ter um único ponto de corte. Uma limitação para uso mais amplo do modelo ordinal é a falta de pacotes estatísticos que o incluam. / Outcomes in epidemiological studies are often measured using ordinal scales, but are nevertheless analyzed through traditional techniques such as logistic regression that require a single dychotomous outcome. Therefore, important information may be lost due to arbitrariness in the choice of a cut-off point. Regression models for ordinal responses have been developed and have the advantages of nr.aking full use of the ordinality of the data, while still providing interpretable estimates for epidemiologists. This paper describes the proportional odds model proposed by McGulla~h and applies it to data from a birth cohort in Southern Brazil. using the statistical package SAS. The objective of this paper was to compare the results of the proportional odds model and of traditional logistic regression using different cut-offs. The outcome was heighUage deficit. expressed in z-scores, with three ordinal leveis. We conclude that the proportional odds model produces an interpretable estimate. similar to an odds ratio, that summarizes the effect over ali leveis of the outcome In the present analysis, the point estimates resulting from ordinal regression were similar to those obtained through traditional logistic regression using a cut-off of -1 z-score. However, ordinal regression lead to smaller confidence intervals (and therefore increased study power), avoiding the arbitrariness of having a single cut-off point. A restriction to the wider use of the ordinal model is the lack of statistical packages that include it.
|
148 |
Regressão de dados binários : distribuição WeibullCaron, Renault 12 March 2010 (has links)
Made available in DSpace on 2016-06-02T20:06:04Z (GMT). No. of bitstreams: 1
2921.pdf: 1787441 bytes, checksum: 0b13c91d9391abb5be3ad333893d026e (MD5)
Previous issue date: 2010-03-12 / Financiadora de Estudos e Projetos / In this work a new class of models for binary data based on Weibull distribution is introduced. A review is made of the most known linkage functions. This class of models has as special case the complementary log-log model and approximates well the logit and probit models. Three real data sets are given to compare the proposed model with many others. In one of these data sets the model is extended to allow multinomial data, that is, a discrete variable with more than two outcomes. The results are very good, because the estimation of parameters is quite simple and the model has shown to be very eficient. / Neste trabalho propõe-se um novo modelo, para conjunto de dados com variável resposta binária, baseado na função densidade acumuladaWeibull. Apresenta- se um resumo das funções de ligação mais conhecidas da literatura. Esta classe de modelos possui como caso especial o modelo complementar log-log e boas aproximações aos modelos logístico e probito. Três conjunto de dados reais são utilizados para comparar o modelo proposto com vários outros modelos. Em um dos conjuntos de dados o modelo _e expandido para suportar variável resposta multinomial, isto _e, variável discreta com mais de dois eventos de interesse. Os resultados obtidos são muito bons, pois a estimação dos parâmetros _e razoavelmente simples e o modelo mostrou-se extremamente eficientes.
|
149 |
Análise de experimentos industriais com respostas categóricas ordenadas : método de Taguchi e modelo de McCullaghVigo, Álvaro January 1994 (has links)
Nas diversas áreas da ciência e na Tecnologia da Qualidade são realizados experimentos para estudar a influência de diversos fatores sobre características (tais como "severidade do desgaste" ou "qualidade de uma solda"), que são registradas como variáveis categóricas ordenadas. Nessa dissertação são estudados aspectos da análise de experimentos com resposta ordinal, como ferramenta da Tecnologia da Qualidade. Os métodos tradicionais de análise estatística (teste Qui-quadrado e testes não-paramétricos, por exemplo) não são eficientes nessa situação, sendo necessários métodos mais elaborados para extrair e interpretar as informações geradas nesses experimentos. Diante disso, foram estudados em detalhe dois métodos alternativos: o primeiro é a técnica da Análise de Acumulação (AA) introduzida por Taguchi (I 987) e, o outro método, baseia-se no modelo de Odds Proporcionais proposto por McCullagh (1980), através do qual é possível estimar e interpretar os efeitos dos fatores. Uma característica especial dessa resposta ordinal é a possibilidade de estar associada a uma variável contínua latente. Assim, os dados observados podem ser vistos como uma categorização dessa variável contínua, não diretamente observável. Essas técnicas de análise são ilustradas para o caso unifatorial e multifatorial, mediante dois conjuntos de dados apresentados na literatura: um estudo observacional e outro experimental. O ajuste dos modelos foi realizado através do procedimento LOGISTIC do pacote estatístico SAS e pelo software ST ATA. Com base nos resultados obtidos, verifica-se a necessidade de modelos paramétricas para estimar magnitude e direção dos efeitos que os fatores provocam na característica de qualidade, uma vez que os métodos tradicionais e a AA não extraem informação suficiente dos dados. Nesse sentido, o modelo de Odds Proporcionais parece um método simples, adequado e eficiente para a análise de dados com resposta ordinal. / In several areas o f Science and Quality o f Technology, experiments are conducted to study the influence of severa! factors on characteristics (as "worn-out severity, or "weld quality"), that are recorded as ordered categorical variables. In this thesis, some aspects o f the analysis o f experiments with ordinal response are studied as a tool of Quality of Technology. The traditional methods of statistical analysis (Chísquare test and non-parametric tests, for example) are not efficient in thís situation, and more sophisticated methods are needed to extract and to interpret the information produced in these experiments. So, two alternative methods are studied in detail: the first is the Accumulation Analysis (AA) technique introduced by Taguchí (1987). The other method is based on the Proportional Odds model proposed by McCullagh (1980), whích allows to estimate and to interpret the effects of the factors. A special characteristics of this response is that it might be associated to a continuous Iatent variable. Thus, the observed data may be seen as a categorization o f this continuous variable not directly observable. These techniques of analysis are illustrated for the single-factor and the multifactor cases, by using two data set presented in the literature: an observational study and an experimental study. The models were adjusted by the procedure LOGISTIC o f the software SAS and by the software ST ATA. The results obtained point to the need of parametric models to estimate the magnitude and direction of the effects caused by the factors in the characteristic of quality, since the traditional methods and the AA do not extract enough information of the data. In thís sense, the Proportional Odds model seems a simple, adequate and efficient method for analyzing data with ordinal response.
|
150 |
A utilização de métodos de regressão para respostas ordinais no estudo de coorte das crianças nascidas em 1993 em Pelotas, RS, BrasilNunes, Luciana Neves January 1999 (has links)
Em estudos epidemiológicos. Os desfechos são freqüentemente medidos em escalas ordinais, entretanto são analisados por técnicas tradicionais. como a reg ressão logística que requer desfecho binário. Consequentemente. informações importantes podem ser perdidas devido a arbitrariedade na escolha do ponto de corte. Modelos de regressão para respostas ordinais têm sido desenvolvidos e têm vantagens de reter a ordinalidade dos dados e fornecer estimativas interpretáveis para os epidemiologistas. Esse artigo descreve o modelo de odds proporcionais proposto por McGullagh e o aplica aos dados da coorte de cnanças nascidas em Pelotas em 1993. usando o pacote estatístico SAS. O objetivo desse artigo foi comparar os resultados do modelo de odds proporcionais e da regressão logística tradicional usando diferentes pontos de corte. O desfecho foi déficit de estatura para idade expressa em escore-z. com três níveis ordinais. Concluiu-se que o modelo de odds proporcionais produz uma estimativa interpretável. similar a uma razão de odds. que sumariza o efeito sobre todos pontos de corte. Na presente análise. as estimativas pontuais resultantes da regressão ordinal foram similares ás obtidas através da regressão logística tradicional usando o ponto de corte do escore-z em -1. Entretanto, a regressão ordinal produziu intervalos de confiança menores (e, portanto, aumento do poder estatístico), evitando a arbitrariedade de ter um único ponto de corte. Uma limitação para uso mais amplo do modelo ordinal é a falta de pacotes estatísticos que o incluam. / Outcomes in epidemiological studies are often measured using ordinal scales, but are nevertheless analyzed through traditional techniques such as logistic regression that require a single dychotomous outcome. Therefore, important information may be lost due to arbitrariness in the choice of a cut-off point. Regression models for ordinal responses have been developed and have the advantages of nr.aking full use of the ordinality of the data, while still providing interpretable estimates for epidemiologists. This paper describes the proportional odds model proposed by McGulla~h and applies it to data from a birth cohort in Southern Brazil. using the statistical package SAS. The objective of this paper was to compare the results of the proportional odds model and of traditional logistic regression using different cut-offs. The outcome was heighUage deficit. expressed in z-scores, with three ordinal leveis. We conclude that the proportional odds model produces an interpretable estimate. similar to an odds ratio, that summarizes the effect over ali leveis of the outcome In the present analysis, the point estimates resulting from ordinal regression were similar to those obtained through traditional logistic regression using a cut-off of -1 z-score. However, ordinal regression lead to smaller confidence intervals (and therefore increased study power), avoiding the arbitrariness of having a single cut-off point. A restriction to the wider use of the ordinal model is the lack of statistical packages that include it.
|
Page generated in 0.0853 seconds