下载文档原格式
客观结构化教学评估（OSTE）在教学查房能力训练中的应用作者：王娜肖砚刘渝松来源：《科教导刊》2023年第33期摘要客观结构化教学评估（OSTE）是一种新型的教学评估与培训的方法。
重庆市中医骨科医院通过设置标准化教学场景，培训标准化学生、患者，运用分站式教学查房实践，并及时进行教学查房反馈的方式，来训练临床带教老师教学查房能力，从接受训练的临床带教老师、标准化学生、教学专家三个维度来评价客观结构化教学评估（OSTE）在带教老师教学查房能力训练中的作用。
关键词客观化结构化教学评估（OSTE）；教学查房；方法中图分类号：G642 文献标识码：A DOI：10.16400/ki.kjdk.2023.33.016The Application of Objective Structured Teaching Evaluation （OSTE） in theTraining of Teaching Ward Rounds AbilityWANG Na， XIAO Yan， LIU Yusong（Science and Education Department， Chongqing Orthopedic Hospital of Traditional Chinese Medicine， Chongqing 400010）Abstract Objective Structured Teaching Evaluation （OSTE） is a new method of teaching evaluation and training. Chongqing Traditional Chinese Medicine Orthopedic Hospital trains standardized students and patients through setting up standardized teaching scenarios， utilizing a step-by-step teaching practice for ward rounds， and providing timely feedback on teaching rounds to train clinical teaching teachers in their ability to conduct ward rounds. This includes training clinical teaching teachers， standardized students， and more Teaching experts evaluate the role of Objective Structured Teaching Evaluation （OSTE） in the training of teaching rounds ability for mentors from three dimensions.Keywords Objective Structured Teaching Evaluation （OSTE）; teaching ward rounds; method現代医院把“医疗、教学、科研”称为促进医院发展的“三驾马车”，2020年国务院办公厅发布了《关于加快医学教育创新发展的指导意见》，指出医学教育是卫生健康事业发展的重要基石，需要夯实高校附属医院医学人才培养主阵地，将医院教学工作提到更高的地位。
How to Read a PaperAugust2,2013S.KeshavDavid R.Cheriton School of Computer Science,University of WaterlooWaterloo,ON,Canadakeshav@uwaterloo.caABSTRACTResearchers spend a great deal of time reading research pa-pers.However,this skill is rarely taught,leading to much wasted effort.This article outlines a practical and efficient three-pass method for reading research papers.I also de-scribe how to use this method to do a literature survey. 1.INTRODUCTIONResearchers must read papers for several reasons:to re-view them for a conference or a class,to keep current in theirfield,or for a literature survey of a newfield.A typi-cal researcher will likely spend hundreds of hours every year reading papers.Learning to efficiently read a paper is a critical but rarely taught skill.Beginning graduate students,therefore,must learn on their own using trial and error.Students waste much effort in the process and are frequently driven to frus-tration.For many years I have used a simple‘three-pass’approach to prevent me from drowning in the details of a paper be-fore getting a bird’s-eye-view.It allows me to estimate the amount of time required to review a set of papers.Moreover, I can adjust the depth of paper evaluation depending on my needs and how much time I have.This paper describes the approach and its use in doing a literature survey.2.THE THREE-PASS APPROACHThe key idea is that you should read the paper in up to three passes,instead of starting at the beginning and plow-ing your way to the end.Each pass accomplishes specific goals and builds upon the previous pass:The first pass gives you a general idea about the paper.The second pass lets you grasp the paper’s content,but not its details.The third pass helps you understand the paper in depth.2.1Thefirst passThefirst pass is a quick scan to get a bird’s-eye view of the paper.You can also decide whether you need to do any more passes.This pass should take aboutfive to ten minutes and consists of the following steps:1.Carefully read the title,abstract,and introduction2.Read the section and sub-section headings,but ignoreeverything else3.Glance at the mathematical content(if any)to deter-mine the underlying theoretical foundations4.Read the conclusions5.Glance over the references,mentally ticking offtheones you’ve already readAt the end of thefirst pass,you should be able to answer thefive Cs:1.Category:What type of paper is this?A measure-ment paper?An analysis of an existing system?A description of a research prototype?2.Context:Which other papers is it related to?Whichtheoretical bases were used to analyze the problem?3.Correctness:Do the assumptions appear to be valid?4.Contributions:What are the paper’s main contribu-tions?5.Clarity:Is the paper well written?Using this information,you may choose not to read fur-ther(and not print it out,thus saving trees).This could be because the paper doesn’t interest you,or you don’t know enough about the area to understand the paper,or that the authors make invalid assumptions.Thefirst pass is ade-quate for papers that aren’t in your research area,but may someday prove relevant.Incidentally,when you write a paper,you can expect most reviewers(and readers)to make only one pass over it.Take care to choose coherent section and sub-section titles and to write concise and comprehensive abstracts.If a reviewer cannot understand the gist after one pass,the paper will likely be rejected;if a reader cannot understand the high-lights of the paper afterfive minutes,the paper will likely never be read.For these reasons,a‘graphical abstract’that summarizes a paper with a single well-chosenfigure is an ex-cellent idea and can be increasingly found in scientific jour-nals.2.2The second passIn the second pass,read the paper with greater care,but ignore details such as proofs.It helps to jot down the key points,or to make comments in the margins,as you read. Dominik Grusemann from Uni Augsburg suggests that you “note down terms you didn’t understand,or questions you may want to ask the author.”If you are acting as a paper referee,these comments will help you when you are writing your review,and to back up your review during the program committee meeting.1.Look carefully at thefigures,diagrams and other illus-trations in the paper.Pay special attention to graphs.Are the axes properly labeled?Are results shown witherror bars,so that conclusions are statistically sig-nificant?Common mistakes like these will separaterushed,shoddy work from the truly excellent.2.Remember to mark relevant unread references for fur-ther reading(this is a good way to learn more aboutthe background of the paper).The second pass should take up to an hour for an expe-rienced reader.After this pass,you should be able to grasp the content of the paper.You should be able to summarize the main thrust of the paper,with supporting evidence,to someone else.This level of detail is appropriate for a paper in which you are interested,but does not lie in your research speciality.Sometimes you won’t understand a paper even at the end of the second pass.This may be because the subject matter is new to you,with unfamiliar terminology and acronyms. Or the authors may use a proof or experimental technique that you don’t understand,so that the bulk of the pa-per is incomprehensible.The paper may be poorly written with unsubstantiated assertions and numerous forward ref-erences.Or it could just be that it’s late at night and you’re tired.You can now choose to:(a)set the paper aside,hoping you don’t need to understand the material to be successful in your career,(b)return to the paper later,perhaps after reading background material or(c)persevere and go on to the third pass.2.3The third passTo fully understand a paper,particularly if you are re-viewer,requires a third pass.The key to the third pass is to attempt to virtually re-implement the paper:that is, making the same assumptions as the authors,re-create the work.By comparing this re-creation with the actual paper, you can easily identify not only a paper’s innovations,but also its hidden failings and assumptions.This pass requires great attention to detail.You should identify and challenge every assumption in every statement. Moreover,you should think about how you yourself would present a particular idea.This comparison of the actual with the virtual lends a sharp insight into the proof and presentation techniques in the paper and you can very likely add this to your repertoire of tools.During this pass,you should also jot down ideas for future work.This pass can take many hours for beginners and more than an hour or two even for an experienced reader.At the end of this pass,you should be able to reconstruct the entire structure of the paper from memory,as well as be able to identify its strong and weak points.In particular,you should be able to pinpoint implicit assumptions,missing citations to relevant work,and potential issues with experimental or analytical techniques.3.DOING A LITERATURE SURVEYPaper reading skills are put to the test in doing a literature survey.This will require you to read tens of papers,perhaps in an unfamiliarfield.What papers should you read?Here is how you can use the three-pass approach to help.First,use an academic search engine such as Google Scholar or CiteSeer and some well-chosen keywords tofind three to five recent highly-cited papers in the area.Do one pass on each paper to get a sense of the work,then read their re-lated work sections.You willfind a thumbnail summary ofthe recent work,and perhaps,if you are lucky,a pointer toa recent survey paper.If you canfind such a survey,youare done.Read the survey,congratulating yourself on your good luck.Otherwise,in the second step,find shared citations and repeated author names in the bibliography.These are thekey papers and researchers in that area.Download the key papers and set them aside.Then go to the websites of thekey researchers and see where they’ve published recently. That will help you identify the top conferences in thatfield because the best researchers usually publish in the top con-ferences.The third step is to go to the website for these top con-ferences and look through their recent proceedings.A quick scan will usually identify recent high-quality related work. These papers,along with the ones you set aside earlier,con-stitute thefirst version of your survey.Make two passes through these papers.If they all cite a key paper that youdid notfind earlier,obtain and read it,iterating as neces-sary.4.RELATED WORKIf you are reading a paper to do a review,you should also read Timothy Roscoe’s paper on“Writing reviews for sys-tems conferences”[3].If you’re planning to write a technical paper,you should refer both to Henning Schulzrinne’s com-prehensive web site[4]and George Whitesides’s excellent overview of the process[5].Finally,Simon Peyton Joneshas a website that covers the entire spectrum of research skills[2].Iain H.McLean of Psychology,Inc.has put together a downloadable‘review matrix’that simplifies paper review-ing using the three-pass approach for papers in experimen-tal psychology[1],which can probably be used,with minor modifications,for papers in other areas.5.ACKNOWLEDGMENTSThefirst version of this document was drafted by my stu-dents:Hossein Falaki,Earl Oliver,and Sumair Ur Rahman.My thanks to them.I also benefited from Christophe Diot’s perceptive comments and Nicole Keshav’s eagle-eyed copy-editing.I would like to make this a living document,updating itas I receive comments.Please take a moment to email meany comments or suggestions for improvement.Thanks to encouraging feedback from many correspondents over the years.6.REFERENCES[1]I.H.McLean,“Literature Review Matrix,”/[2]S.Peyton Jones,“Research Skills,”/en-us/um/people/simonpj/papers/giving-a-talk/giving-a-talk.htm[3]T.Roscoe,“Writing Reviews for Systems Conferences,”http://people.inf.ethz.ch/troscoe/pubs/review-writing.pdf[4]H.Schulzrinne,“Writing Technical Articles,”/∼hgs/etc/writing-style.html[5]G.M.Whitesides,“Whitesides’Group:Writing a Paper,”/∼rlake/Whitesides writing res paper.pdf。
一种三维图像的配准方法成员Paul Jbesl、IEEE，以及NeilD.McKay摘要：本文介绍了一种多方面、表示独立的三维图像的精确计算方法，包括自由曲线和曲面。
该方法处理所有6个自由程度是基于迭代最近点（ICP）算法，这需要一个去找到一个几何实体到一个给定点的最接近点的过程。
ICP算法总是单调收敛到局部的最近平均距离，而经验表明在最初的几次迭代收敛速度快。
因此，给定一组充足的初始旋转和平移为一个特定类的对象具有一定的“图像复杂度”，通过测试每个初始配准可以在全局围内最大限度地减少平均距离的所有六个自由程度。
例如，一个给定的“模型”的图像和感测到的“数据”的图像表示模型的图像的主要部分，它通过测试一个初始的平移和一个相对较小的旋转设置允许给定的模型复杂度来配准几分钟。
这种方法的一个重要应用是配准从不固定的刚性物体与一种理想的形状检验前的几何模型监测到数据。
所描述的方法也是有用的，用于决定的基本问题，如不同的几何表示的重叠（图像等价），以及用于估计未知的点集运动的对应关系。
实验结果表明基于点集，曲线和曲面上的配准算法的能力。
关键词-自由型曲线匹配，自由形态表面匹配，运动估计，姿态估计，四元数，三维配准。
一、引言全局和局部图像匹配度量自由曲线曲面以及点集的匹配，在[ 3 ]中描述了一种试图将计算机视觉中的一个关键问题的描述形式化和统一化的尝试：在传感器坐标系给出的三维数据，它描述了一个数据的图像可能对应一个模型的图像，并给出了在一个模型中的坐标系统中的模型的形状用不同的几何形状表示，估计最佳的旋转和平移对齐，或配准，模型图像和数据的图像距离最小化，从而允许通过一个均方距离度量的等价的形状。
许多应用的关键的利害关系是下面的问题：从一系列图像的分割区域匹配的B样条曲面是在计算机辅助设计（CAD）的一个子集的模式吗？本文提供了一个解决这个自由曲面匹配问题的方案，正如在[ 3 ]和[ 5 ]中定义的一种特殊的情形一样，一个简单的，统一的方法，概括到N 维的，提供的解决方案1)不对应点集匹配问题2)自由曲线的匹配问题。
Unit 1一，Views on language：一、Structural view （language competence）结构主义语言观—The founder：Saussure，lasen freeman&long—The structural view of language sees language as a linguistic system made up of various subsystems：一、the sound system（phonology）二、sound combinations（morphology）the discrete units of meaning 3、the system of combining units of meaning for communication（syntax）—The structural view limits knowing a language to knowing its structural rules andvocabulary2 、Functional view功能主义语言观—Representative：Johnson、marrow、swain canal (the core: grammar)—The function view not only sees language as a linguistic system but also a means for doing things功能不仅以为语言是一个语言系统,但也做情形的一种方式—Learners learn a language in order to be able to doing things with itUse the linguistic structure to express functions3、Interactional view 交互语言观（communicative competence）—Emphasis：appropriateness—Language is a communicative tool，which main use is to build up and maintain social relations between people—Learners need to know the rules for using the language in certain context 二，View on language learning语言学习观1.Process-oriented theories:强调进程are concerned with how the mind organizes new information such as habit formation, induction, making inference, hypothesis testing and generalization.2.Condition-oriented theories: 强调条件emphasize the nature of the human and physical context in which language learning takes place, such as the number of students, the kind of input learners receives, and the atmosphere.3.Behavioristtheory,(Skinner and waston raynor)A the key point of the theory of conditioning is that” you can train an animal to d o anything if you follow a certain procedure which has three major stages, s timulu s, response, and reinforcementB the idea of this method is that language is learned by constant repletion and the reinforcement of the teacher. Mistakes were immediately corrected, and correct utterances were immediately praised.4.Cognitive theory:Chomsky)thinks that language is not a form of behavior,it is an intricate rule-based system a nd a large part of language acquisition is the learni ng of this system.There are a finite number of grammatical rules in the system and with knowledge of these an infinite number of sentences can be produced.5.Constructivist theory:(John Dewey)the constructivist theory believes that lea rning is aproces in which the learner constructs meaning based on his/her own experie nces and what he/her already knows6.Socio-constructivist theory: (Vygotsky) he emphasizes interaction and enga gement with the target language in a social context based on the concept of “Zone of Proximal Development” (ZPD) and scaffolding.Unit 2一，What makes a good language teacher?ethic devotion, professional qualities ，certain desirable personal styles.四， principles of communicative language teaching (CLT) 交际语言教学法原那么1) Communication principle: activities that involve real communication promote l earning.2) Task principle: activities in which language is used for carrying out meaningful tasks promote learning.3) Meaningfulness principle: language that is meaningful to the learner supportsthe learning process.五，Howatt proposes a weak and a strong version of CLT.Weak version: learners first acquire language as a structural system and then lear n how to use it in communication. --- the weak version regards overt teaching of l anguage forms and functions as necessary means for helping learners to develop the ability to use them for communication.Strong version: language is acquired through communication. The learners discov er the structural system in the process of leaning how to communicate.---regards experiences of using the language as the main means or necessary conditions for l earning a language as they provide the experience for learners to see how langua ge is used in communication.六，PPP: presentation，practice，production三. Principles for good lesson planningA. AimB. VarietyC. FlexibilityD. learning abilityE. linkage四. Components of a lesson plan教案的内容A. Background informationB Teaching aimsC. Language contents and skillsD. stages and proceduresE. Teaching aidsF. End of lesson summaryG.. Optional activities and assignmentsH. After lesson reflectionUnit 5二，The role of the teacher 教师的角色1. Controller: control the pace, the time, the target language, the student.2. Assessor: two thingsa. as corrector: correct the mistakes, organizing feed back the learnersb. as evaluator: to create a success-oriented learning, atmosphere, more praise, less criticism3. Organizer : task based on teaching to design tasks and to organize4. Prompter: to give appropriate prompts hints5. Participant: to take part in the activities6. Resource-provider: as a walking dictionaryUnit 6一，Critical Period Hypothesis 关键期假说This hypothesis states that if humans do not learn a foreign language before a certain age ,then due to changes such as maturation of the brain ,it becomes impossible to learn the foreign language like a native speaker.Unit 7三，pennington grammatical pedagogy:1.collocational grammar should biuld on collocational relations between individual lexical items and their subcategories2.Constructive offer learners a way to build elements that can be continually added in sequence3.Contextual it means that elements and structures are taught in relation to their context.四，mechanical practice机械操练1.substitute drills 替换the students substitute a part in a structure so that they getto know how that part function in a sentence2.Transformation drills转换change a given structure in a way so that they are exposed to another similar structureUnit 81. A: passive/receptive words :words that can be recognized or compared inreading and listening but can not be used automatically in speaking and writing.B: active/productive words: words that can be recognized and also be used in speech and writing by learners.Unit 11Sight vocabulary:words that one is able to recognise immediately are often referred to as sight vocabulary.Unit15Testing takes the pencil and paper form and it is usually done at the end of a learning periodAssessmen t involves the collecting of in formation or evidence of a learner s teaching and learning.Evaluation:can be concerned with a whole range of issues in and beyond language education :lessons courses programs and skills can all be evaluated ,四，bloom’s taxonomy 目标分类学1.knowledge知识:recalling facts ,terms,and basic concepts2.prehension明白得:understanding of facts and ideas byorganizing ,comparing,translating interpreting,describing and stating the main ideas3.application运用:applying acquired knowledge,facts ,techniques and rules in a different context.4.analysis分析:identifying relationships,causes or motives,and finding evidence to support main ideas.5.synthesis综合：combing elements in a different way and proposing alternative solutions,creative thinking.6.evaluation 评判：present and defend opinions by making informed judgement about information or ideas based on a set of criteria.、Teaching objectives中心the Ss will be able to understand the main idea of an article about XX and can write a list of XX for XX.辞汇be able to name the new word about XX in english using pictures as cues and be able to tell each other whatXX they like.情感be able to talk about their opinions or feelings about XX to each other.其他tell the five simple forms ofXX can role play the dialogue of XXWarming up.。
AATC C Technical Manual/2014TM 107-2013179Developed in 1962 by AATCC Commit-tee RA23; revised 1967, 1968, 1972,1981, 2009, 2012, 2013; reaffirmed 1975, 1978, 1989, 2007; editorially re-vised 1983, 1985, 1994, 2001, 2005,2008, 2010; editorially revised and re-affirmed 1986, 1991, 1997, 2002. Tech-nically equivalent to ISO 105-E01.1. Purpose and Scope 1.1 This test method is designed to measure the resistance to water of dyed,printed, or otherwise colored textile yarns and fabrics.1.2 Distilled water or deionized water is used in this test method because natural (tap) water is variable in composition.2. Principle2.1 The specimen, backed by multifi-ber test fabric, is immersed in water un-der specified conditions of temperature and time, and then placed between glass or plastic plates under specified condi-tions of pressure, temperature and time.The change in color of the specimen and the staining of the attached multifiber test fabric are observed.3. Terminology 3.1 colorfastness, n.—the resistance of a material to change in any of its color characteristics, to transfer of its color-ant(s) to adjacent materials or both, as a result of the exposure of the material to any environment that might be encoun-tered during the processing, testing, stor-age or use of the material.4. Safety PrecautionsNOTE: These safety precautions are for information purposes only. The pre-cautions are ancillary to the testing proce-dures and are not intended to be all inclu-sive. It is the user’s responsibility to use safe and proper techniques in handling materials in this test method. Manufac-turers MUST be consulted for specific details such as material safety data sheets and other manufacturer’s recommenda-tions. All OSHA standards and rules must also be consulted and followed.4.1 Good laboratory practices should be followed. Wear safety glasses in all laboratory areas.4.2 Manufacturer’s safety recommen-dations should be followed when operat-ing laboratory testing equipment. 4.3 Observe padder safety. Ensure ade-quate guard at the nip point. Normal safe-guards on pad should not be removed.5. Apparatus and Materials (see 12.1)5.1 Perspiration tester (plastic or glass plates are available with the equipment)(see 12.2).5.2 Drying oven—convection.5.3 Multifiber test fabrics (8 mm [0.33in.] bands) containing acetate, cotton,nylon, silk, viscose rayon and wool shall be used for specimens containing silk, or (8mm [0.33 in.] bands) containing ace-tate, cotton, nylon, polyester, acrylic and wool shall be used for specimens with no silk present.5.4 AA TCC 9-Step Chromatic Trans-ference Scale (AATCC Evaluation Proce-dure 8) (see 12.3).5.5 Gray Scale for Color Change (AATCC Evaluation Procedure 1) and Gray Scale for Staining (AA TCC Evalua-tion Procedure 2) (see 12.3).5.6 Wringer.6. Test Solution 6.1 Distilled water or deionized water from an ion-exchange device.7. Test Specimens 7.1 If the specimen to be tested is a fabric, attach a piece of multifiber adja-cent fabric also measuring 5 × 5 ± 0.2 to the specimen measuring 6 × 6 ± 0.2 cm by sewing along one of the shorter sides,with the multifiber fabric next to the face of the specimen.7.2 If the specimen to be tested is a yarn or loose fiber, take a mass of the yarn or loose fiber approximately equal to one half of the combined mass of the adjacent fabrics. Place it between a 5 × 5± 0.2 cm piece of multifiber fabric and a 6 × 6 ± 0.2 cm piece of the non-dyeable fabric, and sew along all four sides.8. Procedure 8.1 Immerse the test specimen in the test solution which is at room tempera-ture with occasional agitation to ensure thorough wetting out (approximately 15min generally required for average fab-rics) (see 12.4).8.2 Remove the test specimen from the test solution and only pass between squeeze rolls (wringer) to remove excess liquor when the wet weight of the test specimen is more than 3 times its dry weight. Whenever possible, the wet weight should be 2.5-3.0 times the dry weight.8.3 Place the test specimen between glass or plastic plates and insert in the specimen unit of the perspiration tester.Adjust the perspiration tester to produce a pressure of 4.5 kg (10.0 lb) on the test specimen (see 12.2).8.4 Heat the loaded specimen unit in an oven at 38 ± 1°C (100 ± 2°F) for 18 h.8.5 Remove the tester from the oven and for each test specimen assembly, sep-arate the multifiber fabric and, if used,the adjacent fabric from the test fabric.Place the multifiber fabric and test fabric specimens separately on a wire screen in a conditioned atmosphere 21 ± 1°C (70 ±2°F) and 65 ± 2% relative humidity over-night.9. Evaluation Method for Color Change 9.1 Evaluate the test specimen for change in color by comparison with the Gray Scale for Color Change (AATCC Evaluation Procedure 1), or by using AATCC Evaluation Procedure 7, Instru-mental Assessment of the Change in Color of a Test Specimen, and record the numerical rating that corresponds to the appropriate one on the Gray Scale.10. Evaluation Method for Staining (see12.7)10.1 Evaluate the staining of the multi-fiber fabric used (see 12.5) by compari-son with the Gray Scale for Staining (AATCC Evaluation Procedure 2), the AATCC 9-Step Chromatic Transference Scale (AA TCC Evaluation Procedure 8),or Instrumental Assessment of Degree of Staining (AATCC Evaluation Procedure 12), and record the numerical rating that corresponds to the appropriate one on ei-ther of them. Report which scale is used (see 12.6).11. Precision and Bias 11.1 Precision . Precision for this test method has not been established. Until a precision statement is generated for this test method, use standard statistical tech-niques in making any comparisons of test results for either within-laboratory or between-laboratory averages.11.2 Bias . Colorfastness to water can be defined only in terms of a test method.There is no independent method for de-termining the true value. As a means of estimating this property, the method has no known bias.AATCC Test Method 107-2013Colorfastness to WaterCopyright © 2013 American Association of Textile Chemists and Colorists180TM 107-2013AATC C Technical Manual/201412. Notes12.1 For potential equipment information pertaining to this test method, please visit the online AATCC Buyer’s Guide at /bg. AATCC provides the pos-sibility of listing equipment and materials sold by its Corporate members, but AATCC does not qualify, or in any way approve, endorse or certify that any of the listed equipment or materials meets the requirements in its test methods.12.2 Horizontal Perspiration Tester: Put all 21 glass or plastic plates into the unit regardless of the number of specimens. After the final glass or plastic plate is put in position on top, set the dual plates with compensating springs in position. Place the 3.6 kg (8.0 lb) weight on top making a total of 4.5 kg (10.0 lb) under the pressure plate. Lock the pressure plate in position by turning the thumb-screws. Remove the weight. Place the unit in the oven on its side,so that the plates and the specimens are vertical.Vertical Perspiration Tester: The plates are held in a vertical position between an indicat-ing scale with a fixed metal plate at one end and an adjustable metal plate at the other end.By means of adjusting screws, the movable plate may be made to exert increasing pressure against the test specimens. When the desired pressure of 4.5 kg (10.0 lb) is indicated on the scale, lock the specimen in it by a set screw.The specimen unit can now be removed from the section applying the pressure. Another specimen unit can be added to the pressure section and the loading procedure repeated.12.3 Available from AATCC, P.O. Box 12215, Research Triangle Park NC 27709; tel:+1.919.549.8141; fax: +1.919.549.8933; e-mail:orders@; web site: .12.4 Or immerse the test specimen in the test solution at room temperature, pass through squeeze rolls (wringer) and reim-merse. Repeat, if necessary, to attain thorough wetting out.12.5 Classify according to the fiber show-ing the greatest stain.12.6 For very critical evaluations and in cases of arbitration, ratings must be based on the Gray Scale for Staining.12.7 CAUTION: It has been reported that the results for staining obtained by this method on fabrics dyed to dark shades (navy,black, etc.) that contain a combination of poly-ester and spandex, or their blends, may not show the full staining propensity of such fab-rics in consumer use. It is, therefore, recom-mended that the staining results obtained by this test not be used for the acceptance testing of such fabrics.Copyright © 2013 American Association of Textile Chemists and Colorists。
麦弗逊前独立悬架汽车的操纵稳定性研究作者：张俊伟学号：0802020407摘要20世纪80年代以来，汽车作为极其重要的交通工具，在交通运输领域和人民日常生活中的地位日益突出。
国内、国际汽车市场的竞争变得空前激烈，用户对汽车安全性、行驶平顺性、操纵稳定性的要求越来越高。
汽车悬架系统是影响车辆动态特性最为关键的子系统，其中由悬架所决定的汽车车轮定位参数对整车操纵动特性有着直接的影响。
悬架的运动学/动力学仿真分析在汽车悬架系统的设计和开发中占有重要的地位。
由于汽车悬架系统是一个复杂的多体系统，其构件之间的运动关系十分复杂，这就给通过传统的计算方法分析悬架的各种特性带来许多的困难。
本论文以机械CAD设计、虚拟样机仿真技术为前题。
提出运用虚拟样机仿真软件ADAMS里的CAR模块分析并进行优化汽车悬架的设计方法。
首先，根据悬架各部件之间的相对运动关系和各部件的参数在ADAMS\CAR中建立某轿车的麦弗逊前悬架的三维CAD模型，再加上路面激励，分析悬架参数在汽车行驶中的变化规律。
然后利用ADAMS\Jnsight对建立的悬架模型进行结构优化，得到悬架系统结构的优化解。
在上述基础上建立了包括前后悬架、发动机、转向系、前后轮胎等在内的整车虚拟样机仿真模型，并根据我国现行整车操纵稳定性试验标准GB/T6323.1.94～GB/T6323.6-94的要求，编写了用于整车操纵稳定性仿真分析的驱动控制文件(DriverControl Files，缩写为DCF)和驱动控制数据文件(DriverControl Da切Rles，缩写为DCD)，进行了转向盘转角阶跃输入试验、转向回正试验、稳态回转试验、蛇行试验和转向轻便性试验等整车操纵稳定性试验仿真分析，并参照GB/T113047-9l《汽车操纵稳定性指标限值与评价方法》对该轿车的操纵稳定性进行了评价计分。
关键词：汽车悬架，建模，ADAMS，操纵稳定性ABSTRACTSince 1980s，the status of automobile has been becoming more and mole outstanding in transportation field and people’s daily lives．The competition of national and intemational automobile markets has become drastic unprecedented，and consumers’demand for safety,handling stability and ride comfort is becoming higher and higher．Automobile suspension system is the most pivotal subsystem that affecting vehicle’s dynamic performances，and the automobile wheel alignment parameters that decide d by suspension has a direct effect to vehicle’s dynamic handling stability．Therefore，the kinematic/dynamic simulation analyses of suspension plays a very important role in suspension's design and exploitation．As suspension system is a complex multi-body system，the movement relation between parts is very complicated，which brings much difficulty for analyzing suspension's performances by traditional calculating methods．Based on mechanical CADdesign and virtual pmtotyping simulation technology, this paper suggested adesign method for analyzing and optimizing vehicle suspension by using virtual prototyping softwareADAMS/CAR．First，build the three—dimensional CAD model of a car’s front Macpherson suspension according to the relative movement relations and parameters of all parts and analyze the suspension parameters’variation rule during driving after adding road actuation．Then optimize the suspension structure and get an optimized result for the uspension system by usingADAMS/Insight．Based 0n the above，the author built the vehicle virtual prototyping simulating model including the front and rear suspensions，the powertrain，the steering system，the front and lear·tires，wrote the driver control files(abbreviation．dco and driver control data files(abbreviation．dcd)for vehicle handling stability simulation analyzing according to the requirements of the current standards GB/T6323．1-94-GB/T6323．6-94 of onr nation’s for vehicle controllability and stability test,carried out simulation and analyses for vehicle handling stability such as steering wheel angle step input test,leturnability test，steady static circular test，pylon course slalom test and steering efforts test，and evaluated the car’s handling stability performance by scoring according to GB/T 13047-91<<Criterion thresholds and evaluation of controllability and stability for automobiles>>．Key words：Automobile suspension，modeling，ADAMS，handling stability0.引言汽车操纵稳定性是指在驾驶者不感到过分紧张、疲劳的条件下，汽车能遵循驾驶者通过汽车转向系给定的方向行驶，且当遭遇外界干扰时，汽车能抵抗干扰而保持稳定行驶的能力。
A communicative approach to writing(交际法写作概念)：To engage them in some act of communication.This means either writing for a specific recipient ,or engaging in an act of creative writing.What is the product-oriented writing?(结果性写作概念)：The production-oriented method of teaching writing pays great attention to the accuracy of the final product but ignores the process.What is the process approach to writing?(过程性写作概念)The process approach to writing does not only pay attention to what students do while they are writing, it also attaches great importance to what they and the teacher do before they start writing and after they finish writing.The main procedures of process writing(过程性写作的步骤)：The main procedures of process writing include creating a motivative to write, brainstorming, mapping, freewriting, outlining ,drafting, editing, revising, proofreading and conferencing. Motivating students to write(激发学生的写作兴趣):1.make the topic of writing as close as possible to students’s life.2.leave students enough room for creativity and imagination.3.prepare students well before writing.4.encourage collaborative group writing as well as individual writing.5.provide opportunities for students to share their writing.6.provide constructive and positive feedback.7.treat students’ errors strategically.8.give students a sense of achievement from time to time.The advantage of e-mail(E-mail写作的好处):1.E-mail provides a prefect mechanism for students to submit drafts and for teachers to look them over at their convenience and send them back with comments.New ideas are shared promptly and can be responded to quickly.2.the teacher can easily store all the drafts of a document for later review and analysis of the revision process.3.send their work to each other simultaneously.An individual students can receive comments from all peers.4.students have a feeling of real-time writing.Two integrated skills（两种综合技能）：Simple integration:a receptive language skill serves as a model for a productive language skill. complex integration:it’s a combination of activities involving different skills, linked thematically. Moral values:Self-control, good health and hygiene, kindness, fairness, self-reliance, sense of duty, reliability, truthfulness, good work attitude, team work, loyalty.The roles of the teacher:Teacher as a role model, teacher as curriculum developer.Assessment purposes(概念):Assessment in ELT means to discover what the leaners know and can do at certain stage of the learning process.Assessment is to find out what the students already know and can do rather than what they do not know and cannot do.Methods for assessment(评价方法)：Summative assessment is mainly based on testing.It is done mostly at the end of a learning periods or the end of a school year.Formative assessment is based on information collected in the classroom during the teaching process for the purposes of improving teaching and learning.The way to gather information(收集信息的方法)：Testing, teacher’s observations, continuous assessment, self-assessment and peer assessment, project work, portfolios.Criteria for assessment(评价的标准)：Criterion-referenced assessment:Criterion-referenced language assessment is based on a fixed standard or a set criterion.Norm-referenced assessment:Norm-referenced assessment is designed to measure how the performance of a particular student or group of students compares with the performance of another students or group of students whose scores are given as the norm.Individual-referenced assessment:Individual-referenced assessment is based on how well the learner is performing relative to his or her own previous performance, or relative to an estimate of his or her individual ability.Different types of learners (学习者的类型)：Visual learners : learn more effectively though the eyes (seeing).Auditory learners: learn more effectively though the ear (hearing).Tactile learners: learn more effectively though touch (hands-on).They learn things by doing. Kinesthetic learners: learn more effectively though body experience.Group learners: learn more effectively though working with others.Individual learners: learn more effectively though working alone.Authority learners: prefer to listen to the teacher more than work with others or work alone. Reflective leaners:learn more effectively when given time to consider options.Multiple-intelligence(多元智能理论):Howard Gardener1.V erbal/Linguistic Intelligence(语言言语智能):The ability to use words effectively, both orally and in writing.2.Musical Intelligence(音乐旋律智能):Sensitivity to rhythm, pitch, and melody.3.Logical/Mathematical Intelligence(逻辑数学智能):The ability to use numbers effectively and reason well.4.Spatial/ Visual Intelligence(空间视觉智能):Sensitivity to form, space, colour, line and shape.5.Bodily/Kinesthetic Intelligence(身体运动智能):The ability to use the body to express ideas and feelings, to solve problems.6.Interpersonal Intelligence(人际关系智能):The ability to understand another person’s mood, feelings, motivations,and intentions.7.Intrapersonal Intelligence(自我认知智能):The ability to understand your strengths, weaknesses, moods, desires, and intentions.8.Naturalist Intelligence(自然观察智能):This is the ability to appreciate the naturalist world and enjoy outdoor activities.The way to learner training/How to help students learn(怎样培养学生的学习能力):1.involve students in an overview of the textbook at the beginning.2.involve students in finding out about themselves.3.introduce students to a number of different learning strategies.4.help leaners set up their own learning goals and make their own plans.5.share lesson aims with students in class and review them by the end of the lesson.e learner diaries as a way to help student reflect on their learning.7.guide students to make plans for learning.e portfolios to promote more autonomous learning.9.help students learn to use resources.How you can create your own resources for teaching and learning (怎样开发隐性资源):e yourself as resourcesing students as resources3.making use of students’ drawings4.making use of surroundings5.shadow theatre6.exploring emotions7.getting students to make their own dictionaries8.letting students to make up their own quizzes9.making batter use of video resourcesing songs for learning11.internet as an important resourceTwo types of evaluation(评价教材的两种方法):On-the-page evaluation:is carried out independent of its users or before it gets into the classroom.In-the-use evaluation: is done based on the users’ opinions i.e. teachers’ as well as learners’, about how useful and effective it is for promoting learning.Two steps of on-the-page evaluation (on-the-page evaluation 的两个步骤):External evaluation:①focus on the external features of a textbook .It is also called macro evaluation.②focus on its internal features ,which maybe termed as micro evaluation.Internal evaluation: intends to investigate the following aspects of a textbook.Features of good textbooks(怎样的教材是好的教材):1.good textbooks should attract the students’ curiosity, interest and attention.2.textbooks should help students feel at ease.3.textbooks should help students develop confidence.4.textbooks should meet students’ needs.5.textbooks should expose students to language in authentic use.6.textbooks should provide students with opportunities to use the target language to achieve communicative purposes.7.textbooks should take into account that the positive effects of language teaching are usually delayed.8.textbooks should take into account that students differ in learning styles.9.textbooks should take into account that students differ in effective factors.10.textbooks should maximise learning potential by encouraging intellectual, aesthetic and emotional involvement which stimulates both right and left brain activities.Suggestions for adapting materials(调整教材的方法):1.Adding2.Deleting or omitting3.Modifying4.Simplification5.ReorderingThree steps of textbook adaption(调整步骤):The first step is macro adaptation ,which is ideally done before the language programme begins. The second step of adaptation is adapting a unit.The third step is adaptation of specific activities in a unit.。
EVALUATION OF A THREE-STEP METHODFOR CHOOSING THE NUMBER OF BOOTSTRAP REPETITIONSBYDONALD W. K. ANDREWS AND MOSHE BUCHINSKYCOWLES FOUNDATION PAPER NO. 1125COWLES FOUNDATION FOR RESEARCH IN ECONOMICSYALE UNIVERSITYBox 208281New Haven, Connecticut 06520-82812006/Journal of Econometrics103(2001)345–386/locate/econbaseEvaluation of a three-step method for choosing the number of bootstrap repetitionsDonald W.K.Andrews a,Moshe Buchinsky b;c;d;∗a Cowles Foundation for Research in Economics,Yale University,New Haven,CT06520-8281,USAb Department of Economics,Brown University,Box B,Providence,RI02912,USAc National Bureau of Economic Research,1050Massachusetts Avenue,Cambridge,MA02138,USAd INSEE-CREST,15Boulevard Gabriel PÃe ri,92245Malako Cedex,FranceAccepted24October2000AbstractThis paper provides a variety of Monte Carlo simulations that evaluate theÿnite-sample performance of the three-step method for choosing the number of boot-strap repetitions,suggested by Andrews and Buchinsky(Econometrica67(2000) 23–51).The simulations cover bootstrap standard errors,conÿdence intervals,tests, and p-values.Three commonly used econometric applications are considered:linear regression,binary probit,and quantile regression.In brief,weÿnd that the three-step method works very well in all of the contexts examined here.We alsoÿnd that the number of bootstrap repetitions commonly used in econometric applications is much less than needed to achieve accurate bootstrap quantities.?2001Elsevier Science S.A.All rights reserved.JEL classiÿcation:C12;C13;C14;C15Keywords:Bootstrap;Bootstrap repetitions;Coe cient of excess kurtosis;Conÿdence interval;Density estimation;Hypothesis test;p-value;Quantile;Simulation;Standard error estimate∗Corresponding author.Tel.:+1-401-863-2951;fax:+1-401-863-1970.E-mail address:moshe buchinsky@(M.Buchinsky).0304-4076/01/$-see front matter?2001Elsevier Science S.A.All rights reserved.PII:S0304-4076(01)00047-1346 D.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–3861.IntroductionAndrews and Buchinsky(2000)consider the problem of choosing the num-ber of bootstrap repetitions B for a wide variety of bootstrap procedures.They introduce a three-step method for doing so.This method is designed to ad-dress the problem that one can obtain a‘di erent answer’from the same data merely by using di erent simulation draws if B is too small,but computa-tional costs can be great if B is chosen to be extremely large.The three-step method of Andrews and Buchinsky(2000)determines B to attain a spec-iÿed level of accuracy.In consequence,one can obtain accurate bootstrap quantities with the minimum computational e ort.The method is justiÿed by asymptotics as B→∞.The primary purpose of this paper is to investigate theÿnite sample prop-erties of the three-step method.We address the question of whether the three-step method delivers the desired level of accuracy inÿnite samples.A secondary purpose of this paper,independent of the three-step method,is to determine the magnitudes ofB necessary to obtain di erent levels of ac-curacy in a variety of bootstrap situations.The results let one judge whether typical choices for B used in the literature are appropriate.We investigate theÿnite sample properties of the three-step method in a variety of di erent contexts.We consider bootstrap standard error estimates, symmetric two-sided conÿdence intervals,tests with a given signiÿcance level ,and p-values.We consider these bootstrap applications in a linear regres-sion model,a binary probit model,and a quantile regression model.In each model,the observations are independent and identically distributed(iid)and the sample size is taken to be small,only25observations.In all cases,we consider the standard nonparametric bootstrap based on the empirical distri-bution function.The measure of‘accuracy’used by the three-step method is the percentage deviation of the bootstrap quantity of interest based on B bootstrap repetitions, from the ideal bootstrap quantity for which B=∞.For the four bootstrap applications considered here,the bootstrap‘quantities of interest’are the stan-dard error estimate,the length of the conÿdence interval,the critical value of the test,and the p-value.For example,for standard error estimates,accuracy is measured in terms of the percentage deviation of the bootstrap standard er-ror estimate for a given(ÿnite)value of B,from the ideal bootstrap standard error estimate.The percentage deviation of any bootstrap quantity for a given value of B is stochastic,because the bootstrap simulations are random.To determine a suitable value of B,we specify a bound on the relevant percentage deviation, denoted pdb,and we require that the actual percentage deviation be less than this bound with a speciÿed probability,1− ,close to one.The three-step method takes pdb and as given and provides a data-dependent method toD.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–386347 determine a value of B,denoted B∗,to obtain the desired level of accuracy. Three steps are required because the relevant features of the problem need to be determined in the initial two steps before it is possible to determine a suitable choice of B in the third step.In the simulations,we assess the precision of the three-step method as follows.For each simulation,we calculate whether the actual percentage de-viation of the bootstrap quantity based on the value of B selected via the three-step method is less than the percentage deviation bound pdb.Then,we compare the fraction of cases over all of the Monte Carlo simulations where this is true,denoted the empirical level,with the nominal level1− .The three-step method performs well if the empirical level is close to1− . The results indicate that in most cases the empirical levels are quite close to the nominal levels.For example,for(pdb; )=(10;0:05),we obtain em-pirical levels of0.947,0.949,and0.942for bootstrap standard error estimates in the linear regression model with errors with t distribution withÿve degrees of freedom(t5),the binary probit model,and the quantile regression model with t5errors,respectively,in comparison to the nominal level of0.950.For symmetric90%conÿdence intervals,the corresponding empirical levels are 0.958,0.958,and0.958.For tests with signiÿcance level0.05,the correspond-ing empirical levels are0.940,0.947,and0.942,respectively.For p-values with p=0:10,the corresponding empirical levels are0.951,0.944,and0.947. In general,the empirical levels for the linear and quantile regression mod-els with normal,rather than t5,errors are even closer to the0.950nominal level.The simulation results show that the precision of the three-step method does not vary greatly across the di erent models considered.The results also show that the precision varies somewhat across the di erent type of bootstrap application considered,with standard errors being the best and p-values being the worst,but that the variation is not too great.The simulation results clearly indicate that the precision of the three-step method depends primarily on how tight the(pdb; )bound is.The smaller the values of pdb and ,the greater is the required number of bootstrap repetitions B∗,and the greater is the precision of the three-step method.The reason is that the three-step method is based on asymptotics as B→∞. Overall,we conclude that the three-step method works very well over the range of bootstrap applications and models that are considered in the simulations.We note that the three-step method is applicable in numerous cases that are not considered in this paper.It applies to bootstrap equal-tailed percentile t conÿdence intervals,one-sided percentile t conÿdence intervals,conÿdence regions,and bias-correction.It applies to parametric and semiparametric boot-straps for iid and temporally dependent samples,to residual-based regres-sion bootstraps,as well as to nonparametric block bootstraps for temporally348 D.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–386 dependent samples.See Andrews and Buchinsky(2000)for details.For want of time and space,we do not consider these cases here.Next,we discuss the magnitude of the B values that are needed to obtain accurate bootstrap quantities.In the econometrics literature,it is common for 100or so bootstrap repetitions to be used.A number of this magnitude is noticeably smaller than the numbers obtained in the simulations.For example, for(pdb; )=(10;0:05),we obtain the following median values(over the simulations)of the B∗values selected by the three-step method for the same models as discussed above:287,207,and291for bootstrap standard error estimates,511,409,and543for symmetric90%conÿdence intervals,767, 828,and834for tests with signiÿcance level0.05,and3580,3690,and3622 for p-values with p=0:10.Note that these median B∗values are very good indicators of the median values of B that are necessary to obtain a(pdb; ) accuracy of exactly(10;0:95),because the empirical levels of the three-step method are quite close to the nominal level of0.95.These median B∗values vary considerably across the di erent bootstrap applications considered and with the speciÿed degree of accuracy(pdb; ) within each application.They also vary somewhat across the di erent models considered.In the case of p-values,the level of accuracy given by(pdb; )= (10;0:05)may be more than one requires.In this case the large values of B given above would be replaced by smaller values,when larger(pdb; )values are speciÿed.Nevertheless,the results indicate that if the speciÿed level of accuracy is desired,then the number of bootstrap repetitions required can be quite large.We conclude that to obtain results that do not depend on the particular bootstrap simulation draws employed,one needs to use more bootstrap repe-titions than is commonly used in the econometrics literature.How many more depends on the type of bootstrap application,the model under consideration, and the desired level of accuracy.Papers in the literature that are related to the three-step method considered here include Efron and Tibshirani(1986),Hall(1986),Davison and Hink-ley(1997,Sections2:5:2and4:2:5),Davidson and MacKinnon(2000),and Andrews and Buchinsky(2000,2001).Efron and Tibshirani(1986,Section9)provide a simple formula that relates the coe cient of variation of the bootstrap standard error estimator,as an estimate of the true standard error,to the coe cient of variation of the ideal bootstrap standard error estimator,as an estimate of the true standard error. Their formula depends on some unknown parameters that are not estimable. Hence,Efron and Tibshirani use their formula to suggest a range of plausible values of B,rather than a speciÿc value of B.Hall(1986)considers unconditional coverage probabilities of conÿdence intervals,i.e.,coverage probabilities with respect to the randomness in the data and the bootstrap simulations.The three-step method considered hereD.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–386349 focuses on conditional coverage probabilities,i.e.,coverage probabilities with respect to the randomness in the data conditional on the bootstrap simulations. The reason is that one does not want to be able to obtain‘di erent answers’from the same data due to the use of di erent simulation draws.Davison and Hinkley(1997,Section2:5:2)provide formulae that decom-pose the variance of bootstrap bias correction estimates,variance estimates, and quantile estimates into the part that is due to simulation and the part that is due to sample variation.They use these formulae to suggest values of B. Davison and Hinkley(1997,Section4:5:2)provide some formulae for the e ect of B on the power of a test.Davidson and MacKinnon(2000)propose a pretesting method of choosing B for a test with a given signiÿcance level that aims to ensure that the probability is small that there is a di erence between the conclusions of the ideal bootstrap test and the bootstrap test based on B bootstrap repetitions. In contrast,the three-step method aims to achieve a bootstrap test that has good conditional signiÿcance level and power conditional on the simulation randomness by determining an accurate critical value.Andrews and Buchinsky(2001)provides a three-step method for choosing B for the BC a conÿdence intervals of Efron(1987).The method is analogous to the three-step method considered here for percentile t conÿdence intervals. The remainder of this paper is organized as follows.Section2provides the notation and describes the bootstrap applications of interest,viz.,standard er-ror estimates,symmetric two-sided percentile t conÿdence intervals,tests for a given signiÿcance level ,and p-values.Section3outlines the three-step method.It also provides a number of tables that illustrate the magnitudes of various quantities that enter the calculations in the three-step method. Section4explains the design of the Monte Carlo experiments.Section5 provides the results of the Monte Carlo experiment.Section6provides a brief summary and concluding remarks.2.Notation and description of the bootstrap applicationsIn this section,we present the notation used throughout the paper and introduce the bootstrap applications considered in the paper,viz.,standard error estimates,conÿdence intervals,tests for a given signiÿcance level,and p-values.2.1.NotationFirst,we outline the general framework.Suppose that we are interested in some unknown quantity .For example, could be an exact standard error,conÿdence interval length,critical value,or p-value.We would like to350 D.W.K.Andrews,M.Buchinsky /Journal of Econometrics 103(2001)345–386estimate using an ‘ideal’bootstrap estimate,denoted ˆ ∞.Analytic calcula-tion of ˆ∞is intractable in most cases,so we use an estimate ˆ B of ˆ ∞that is based on a ÿnite number,B ,of bootstrap simulations.The three-step method of Andrews and Buchinsky (2000)speciÿes a data-dependent method of se-lecting B such that ˆB is close to ˆ ∞within a prespeciÿed level of accuracy.We describe this method below.The observed data are a sample of size n :X =(X 1;:::;X n ) .Let X ∗=(X ∗1;:::;X ∗n ) be a bootstrap sample of size n based on the original sample X .In this paper,we consider the case where X is a sample of iid random vectors and the bootstrap sample X ∗is an iid sample drawn from the empirical distribution ˆF (i.e.,a random sample of size n drawn from the original sample with replacement).This is the most commonly used bootstrap.Let {X ∗b :b =1;:::;B }denote B iid bootstrap samples,each with the same distribution as X ∗.All probability statements and the probability and expectation operators P ∗and E ∗,respectively,refer to the randomness in the iid bootstrap samples {X ∗b :b =1;:::;B }conditional on the observed data X .The accuracy of ˆ B is measured by the percentage deviation of ˆ B from ˆ∞:100|ˆ B −ˆ ∞|ˆ∞:(1)This percentage deviation is random conditional on the sample X ,because it depends on the random bootstrap simulations that are used to calculate ˆ B .Let 1− denote a probability close to one,such as 0.95.Let pdb be a bound on the percentage deviation of ˆB from ˆ ∞.The three-step method of Andrews and Buchinsky (2000)is designed to determine B such that P ∗ 100|ˆ B −ˆ ∞|ˆ ∞6pdb ≈1− ;(2)where ≈denotes ‘is approximately equal to’.The three-step method is based on the following asymptotic result:B 1=2(ˆ B −ˆ ∞)=ˆ ∞→d N (0;!)as B →∞;(3)where the asymptotic variance !depends on the particular application con-sidered.1The three-step method depends on an estimator ˆ!B of !.This estimator is based on the bootstrap samples {X ∗b :b =1;:::;B }.1This result holds with probability one with respect to the distribution of the original sample.In the examples in which ˆB is a sample quantile,viz.,the conÿdence interval and test for a given signiÿcance level examples,this result holds as both B →∞and n →∞and it holds with probability one with respect to the distribution of the inÿnite sequence of random variables that yields the original samples for di erent values of n .D.W.K.Andrews,M.Buchinsky /Journal of Econometrics 103(2001)345–386351In the following subsections,we specify the quantities ;ˆ∞;ˆ B ;!,and ˆ!B in each of the applications of interest.2.2.Standard errorsLet ˆÂ=ˆÂ(X )be an estimator of a scalar parameter Â0based on the sample X .For standard error estimates,the quantity is the standard error,se ,of ˆÂ:se =(E (ˆÂ(X )−E ˆÂ(X ))2)1=2;(4)where E denotes expectation with respect to the randomness in X .Let ˆÂ∗b =ˆÂ(X ∗b )for b =1;:::;B denote B bootstrap estimates of Â0.The ‘ideal’bootstrap standard error estimator of se is given byse ∞=(E ∗(ˆÂ∗b −E ∗ˆÂ∗b )2)1=2:(5)The bootstrap standard error estimator based on B bootstrap repetitions is se B =⎛⎝1B −1B b =1 ˆÂ∗b −1B B c =1ˆÂ∗c2⎞⎠1=2:(6)In this case,ˆ ∞= se ∞and ˆ B = se B .Provided E ∗((ˆÂ∗b )2)¡∞;lim B →∞ se B = se ∞almost surely by the law of large numbers.In this application,the variance !of (3)depends on the coe cient ofexcess kurtosis ,denoted 2,of the bootstrap estimator ˆÂ∗b .2In particular,!=(2+ 2)=4;where 2=E ∗(ˆÂ∗b − )4= se 4∞−3and =E ∗ˆÂ∗b :(7)A consistent estimator of !isˆ!B =(2+ˆ2B )=4;where ˆ 2B =1B −1B b =1(ˆÂ∗b −ˆ B )4= se 4B −3and ˆ B =1B B b =1ˆÂ∗b :(8)By the law of large numbers and Slutsky’s Theorem,it follows that lim B →∞ˆ B = ;lim B →∞ˆ 2B = 2,and lim B →∞ˆ!B =!almost surely,provided se ∞=0and E ∗((ˆÂ∗b )4)¡∞.The estimator ˆ!B tends to be biased toward zero in small samples,so we also consider the bootstrap bias-corrected version of ˆ!B as an estimator of !.The iid sample of B bootstrap estimates of Â0is ∗B=(ˆÂ∗1;:::;ˆÂ∗B ).For 2If ˆÂ∗b has a normal distribution then 2=0,if ˆÂ∗b has kurtosis greater than that of a normal distribution then 2¿0,and 2¡0otherwise.352 D.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–386 present purposes,we think of(ˆÂ∗1;:::;ˆÂ∗B)as being the original sample andˆ 2B as being an estimator based on this sample that we want to bootstrap bias correct.LetˆG denote the empirical distribution of(ˆÂ∗1;:::;ˆÂ∗B).Consider R independent bootstrap samples{ ∗∗Br:r=1;:::;R},where each bootstrap sample ∗∗Br=(ˆÂ∗∗1r;:::;ˆÂ∗∗Br)is a random sample of size B drawn fromˆG.The bootstrap bias-corrected estimatorˆ 2BR of 2for R bootstrap repetitions isˆ 2BR=2ˆ 2B−1RRr=1ˆ 2( ∗∗Br);whereˆ 2( ∗∗Br)=[1=(B−1)]Bb=1(ˆÂ∗∗br−(1=B)Bc=1ˆÂ∗∗cr)4([1=(B−1)]Bb=1(ˆÂ∗∗br−(1=B)Bc=1ˆÂ∗∗cr)2)2−3:(9)2.3.Symmetric two-sided percentile t conÿdence intervalsNext,we consider symmetric two-sided percentile t conÿdence intervals for the scalar parameterÂ0.These intervals are symmetric about the estimatorˆÂ. In the models that we consider below,the normalized estimator n1=2(ˆÂ−Â0) has an asymptotic normal distribution as n→∞.Letˆ =ˆ (X)denote a consistent estimator of the asymptotic standard error of n1=2(ˆÂ−Â0).Let T=|n1=2(ˆÂ−Â0)=ˆ |:(10)Let q01− denote the1− quantile of T.The‘theoretical’symmetric two-sidedpercentile t conÿdence interval with exact conÿdence level100(1− )%is J SY=[ˆÂ−n−1=2ˆ q01− ;ˆÂ+n−1=2ˆ q01− ]:(11)The quantity of interest in this case is q01− ,which is proportional to thelength of the conÿdence interval.Deÿneˆ ∗b=ˆ (X∗b )and T∗b=|n1=2(ˆÂ∗b−ˆÂ)=ˆ ∗b|for b=1;:::;B.The1−quantile of T∗b ,denotedˆq1− ;∞,is the ideal bootstrap estimate of q01−.Thus,ˆ ∞equalsˆq1− ;∞in this application.The ideal bootstrap symmetricpercentile t conÿdence interval of approximate conÿdence level100(1− )%isˆJ SY;∞=[ˆÂ−n−1=2ˆ ˆq1− ;∞;ˆÂ+n−1=2ˆ ˆq1− ;∞].3Letˆq1− ;Bdenote the1− sample quantile of the B bootstrap t statistics {T∗b:b=1;:::;B}(deÿned more precisely below).In this application,ˆ Bequalsˆq1− ;B.The bootstrap symmetric percentile t conÿdence interval of3The conÿdence level of this bootstrap conÿdence interval exhibits higher order improvements over the corresponding conÿdence level based on the delta method;e.g.,Beran(1988)and Hall (1992).D.W.K.Andrews,M.Buchinsky/Journal of Econometrics103(2001)345–386353 approximate conÿdence level100(1− )%based on B bootstrap repetitions isˆJ SY;B =[ˆÂ−n−1=2ˆ ˆq1− ;B;ˆÂ+n−1=2ˆ ˆq1− ;B]:(12)Following Hall(1992,p.307),for this application,we choose B so that =(B+1)=1− for some positive integer .We consider values of that are rational and can be written as= 1= 2(13) for some positive integers 1and 2(with no common integer divisors).Then, B= 2h−1and =( 2− 1)h for some positive integer h.Let{T∗B;b:b=1;:::;B} denote the ordered sample of bootstrap T statistics.Then,for B and asabove,the1− sample quantileˆq1− ;B of{T∗b:b=1;:::;B}isˆq1− ;B=T∗B; :(14)That is,ˆq1− ;B is the th order statistic of{T∗b:b=1;:::;B}.In this application,!is given by!= (1− )=(4z21− =22(z1−a=2));(15) where z1− =2and (·)denote the1− =2quantile and the density function, respectively,of the standard normal distribution.The estimateˆ!B isˆ!B= (1− )(1=ˆg B)2=ˆq21− ;B;where1=ˆg B=BB(T∗B; +ˆmB−T∗B; −ˆmB); =(B+1)(1− );ˆm B=int(c B2=3)and c =6z21− =22(z1− =2)2z21− =2+11=3:(16)Note that1=ˆg B is Siddiqui’s(1960)estimator of the reciprocal of the densityof T∗b with a plug-in estimator of the bandwidth parameter,viz.,ˆm B.4Symmetric conÿdence intervals are appropriate only if the distribution of the t statistic upon which the conÿdence interval is based has a distribution that is approximately symmetric.The asymptotic distribution of the t statistic is normal in most cases.So,in large samples,its distribution is approximately symmetric.In small samples,however,its distribution may not be approxi-mately symmetric.If the t statistic has a noticeably asymmetric distribution, then a symmetric conÿdence interval may be misleading.In such a case, an equal-tailed two-sided conÿdence interval is more appropriate.Andrews and Buchinsky(2000)describe a three-step for choosing B for equal-tailed percentile t conÿdence intervals.Andrews and Buchinsky(2001)describe 4This estimator has been analyzed by Bloch and Gastwirth(1968)and Hall and Sheather (1988).a three-step for choosing B for the equal-tailed BC a conÿdence intervals of Efron (1987).2.4.Tests for a given signiÿcance levelHere,we consider one-sided tests for a given signiÿcance level .The null and alternative hypotheses areH 0:Â0=0andH 1:Â0¿0:(17)The test statistic considered in this case isT =n 1=2(ˆÂ−Â0)=ˆ ;(18)where ˆÂand ˆ are deÿned as above.The ‘theoretical’test of exact signiÿcancelevel rejects the null hypothesis if T ¿q 01− ,where q 01− is the 1− quantileof T under the null hypothesis.The quantity of interest in this case is q 01− .The bootstrap version of the test statistic depends on the type of resam-pling used to construct the bootstrap samples.If the bootstrap samples are generated by a method that does not impose the null hypothesis,such asthe nonparametric bootstrap,then T ∗b =n 1=2(ˆÂ∗b −ˆÂ)=ˆ∗b .On the other hand,if the bootstrap samples are generated by a method that imposes the null hypoth-esis,such as the parametric bootstrap based on Â0=0,then T ∗b =n 1=2(ˆÂ∗b −0)=ˆ∗b .Let ˆq 1− ;∞denote the 1− quantile of T ∗b .The ideal bootstrap test of ap-proximate signiÿcance level rejects the null hypothesis if T ¿ˆq 1− ;∞.Theestimate ˆ ∞in this case is ˆq 1− ;∞.Let ˆq 1− ;B denote the 1− quantile of {T ∗b:b =1;:::;B }.We take B and as in the previous subsection.Thus,ˆq 1− ;B equals T ∗B;,the th order statistic of {T ∗b :b =1;:::;B }.The bootstrap test of approximate signiÿcance level based on B bootstrap repetitions rejects the null hypothesis ifT ¿ˆq 1− ;B :(19)In this case,ˆ B is ˆq 1− ;B .The quantity !in this example is!= (1− )=(z 21− 2(z 1− )):(20)The estimator ˆ!B is the same as in (16),but with c deÿned byc = 1:5z 21− =22(z 1− )1− 1=3:(21)2.5.p -valuesWe now consider a testing problem in which one wants to report a p -value.In this case,the quantity of interest is the exact p -value.The null andalternative hypotheses are as in(17).Let T and T∗b be deÿned as in theprevious subsection.The ideal bootstrap p-value and the bootstrap p-value based on B bootstrap repetitions areˆp∞=P∗(T∗b¿T)andˆpB =1BBb=11(T∗b¿T):(22)In this case,ˆ ∞=ˆp∞andˆ B=ˆpB .We assume thatˆp∞does not equal zeroor one.The variance!and its estimateˆ!B are given by!=(1−ˆp∞)=ˆp∞andˆ!B=(1−ˆpB )=ˆpB:(23)3.A three-step method of determining BWe now specify the three-step method of Andrews and Buchinsky(2000a) for determining B to achieve a desired accuracy ofˆ B for estimatingˆ ∞. Recall that the desired accuracy is speciÿed by a(pdb; )combination.3.1.The methodThe three-step method depends on a preliminary estimate!1of the asymp-totic variance!of B1=2(ˆ B−ˆ ∞)=ˆ ∞.For the applications of Section2,we use the following:Standard errors:!1=1=2;Symmetric two-sided conÿdence intervals:!1= (1− )=(4z21− =2 2(z1− =2));Tests for a given signiÿcance level:!1= (1− )=(z21− 2(z1− ));p-values:!1= (T)=(1− (T));(24)where z1− ; (·),and (·)denote the1− quantile,density function,and distribution function,respectively,of the standard normal distribution.5These speciÿcations of!1are based on asymptotics,but the three-step method is 5The last three formulae for!1in(24)and the corresponding formulae for!given above are suitable only when T has an absolute standard normal,standard normal,and standard normal asymptotic distribution,respectively,which is the case considered here.Andrews and Buchinsky (2000)give the appropriate formulae for the general case,which includes the common testing case in which T has an asymptotic chi-squared distribution.not too sensitive to their choice,because it uses a ÿnite sample estimate of !in the last step.Let int(a )denote the smallest integer greater than or equal to a .The three-step method is as follows:Step 1.Given !1,computeB 1=int (10;000z 21− =2!1=pdb 2)(25)or,if ˆB is a 1− sample quantile,compute B 1= 2h 1−1and 1=(B 1+1)(1− ),where = 1= 2and h 1=int (10;000z 21− =2!1=(pdb 22)).Step 2.Simulate B 1bootstrap samples {X ∗b :b =1;:::;B 1}and compute an improved estimate ˆ!B 1of !using the appropriate formulae given in (8),(16),(21),or (23),with B replaced by B 1.Step puteB 2=int (10;000z 21− =2ˆ!B 1=pdb 2)(26)or,if ˆB is a 1− sample quantile,compute B 2= 2h 2−1,where h 2=int (10;000z 21− =2ˆ!B 1=(pdb 2 2)).Take the desired number of bootstrap repeti-tions to be B ∗=max {B 2;B 1}.3.2.Justiÿcation of the three-step methodThe justiÿcation of the three-step method is that as pdb →0(and n →∞when ˆ B is a sample quantile),we have P ∗ 100|ˆ B 2−ˆ ∞|ˆ ∞6pdb→1− :(27)Note that B 2depends on pdb in (27)via (26)and B 2→∞as pdb →0.Eq.(27)implies that the three-step method attains precisely the speci-ÿed accuracy asymptotically using ‘small pdb ’asymptotics when !¿!1.If !¡!1,then B ∗=B 1¿B 2with probability that goes to one as pdb →0(andn →∞when ˆB is a sample quantile)and the accuracy of ˆ B ∗for approxi-mating ˆ∞exceeds that of (pdb; ).This is a consequence of the fact that it would be silly to throw away the extra B 1−B 2bootstrap estimates that have already been calculated in Step 2.Because one normally speciÿes a small value of pdb ,the asymptotic re-sult (27)should be indicative of the relevant non-zero pdb behavior of the three-step method.The simulation results of Section 5are designed to exam-ine this.We note that the asymptotics used here are completely analogous to large sample size asymptotics with pdb driving B 2to inÿnity as pdb →0and B 2playing the role of the sample size.For more details on the asymptotic justiÿcation,see Andrews and Buchin-sky (2000).。
