1、科学研究的基本要素,创造性的因素 创新性或革新的因素 科学方法的应用 新知识的产生和应用,Scientific Research Methodology,Scientific research methodology is a procedure used to logically develop an understanding of the world in which we live.,The basic steps involved in a scientific study:,Define an area, subject, question, or problem to be st
2、udied. Do background research to determine what is already known about the subject to be studied. Develop a hypothesis or multiple hypotheses to guide the project or study.,The basic steps involved in a scientific study: (continued),4. Gather data and observations by conducting field studies or expe
3、riments. 5. Analysis and interpret results and data. 6. Use the analysis to support, reject, or reevaluate the hypothesis and the null hypothesis. 7. Publish the results of the scientific research.,科学研究的基本方法,观察法：临床观察 实验法：实验条件下的观察 调查法：流行病学调查、营养调查,Research may be classified according to purpose as: 1.
4、 Descriptive studies 2. Analytical studies,Descriptive studies include: 1. Qualitative research generate narrative data, i.e., data described in words instead of numbers. 2. Case reports 3. Case series 4. Survey research,Analytical studies include: 1. Experimental designs, such as clinical trials; 2
5、. Observational research designs, such as cohort or follow-up studies and case control studies.,Analytical studies: 1. Analytical studies are designed to test hypotheses concerning the effects of specific factors of interest and allow causal associations to be detected. 2. Only experimental studies
6、in which intervention is designed by the researchers provide proof of cause and effect.,科研选题,是否具备科研的基本要素(meaningful and important) 可行性：科研条件与科研环境(feasible, start with a simple question) 科研经费 成本与效益 科研课题的优先性 个人兴趣 时间,科研设计,制定科研课题研究技术方案和计划实施方案 做什么、怎么做、何时做、谁来做 科研设计包括专业设计和统计学设计（实验设计）,专业设计,技术路线 简图（框架图） 实验观察内
7、容、指标、方法 实验对象 动物、人 被试因素 单因素、多因素；独立、交叉 观测指标 关联、灵敏、稳定、特异、精确、可行,实验设计（统计学设计）,以数理统计学理论技术为基础 在专业设计的基础上 进行实验设计 对实验数据进行数理统计分析,实验设计基本要素,受试对象 处理因素1. 主要因素2. 非处理因素3. 处理因素的标准化4. 处理因素的水平 实验效应,实验设计的基本原理,实验组 VS 对照组T + S e + s O + S 0 + s 实验 I 组 VS 实验 II 组 T1 + S1 e1 + s1T2 + S2 e2 + s2,实验设计的基本原则,对照原则：体现相对性及控制非实验因素和实
8、验误差 均衡原则：控制非实验因素和实验误差 随机原则：控制非实验因素和实验误差 重复原则：控制非实验因素和实验误差,实验对照,空白对照 对照组不施加任何处理因素 实验对照 对照组施加部分实验因素，但不是所研究的处理因素 标准对照 用标准方法做对照 历史对照 以过去的研究结果做对照 自身对照 对照与实验在同一受试对象进行 相互对照 几种处理（或水平）互为对照,均衡化方法,交叉均衡 在实验单元中设立实验组和对照组，使两组的非处理因素一致。分层均衡 用分层方法使各处理组中的非处理因素得到均衡。它是将非处理因素按不同水平划分为若干单元组，然后在每个单元组内安排处理因素。,随机化方法,每个受试对象接受每
9、个处理的机会完全 相等。随机化方法有：摸球或抽签法 随机数字表法 随机化分组表法,重复原则与样本含量的估计,重复程度表现为样本含量的大小。重复次数越多样本含量越大，实验结果越准确。最低样本含量？ ( = 0.05, = 0.10, 已知，实验效应的组间差值？),第一类误差的概率()，即检验水准。 越小，所需样本例数越多。 检验效能(1-)，即在特定的水准下，若总体间确实存在差别，该次试验能发现此差别的概率。为第二误差的概率越大，所需样本例数越多。 结合和的大小可以这样理解，即用如此确定的样本例数作实验，若总体参数间确实相差这么大，则预期有1-的概率(把握度)按检验水准得出有差别的结论。,常用实
10、验设计方法,单组比较（自身对照）设计 对每个观察单位进行两次观察，第一次不给予处理，第二次给予处理，比较两次观察的结果。 第一次是A处理，第二次是B处理。 将受试对象随机分为两组。第一组给予A处理，第二组给予B处理，观察结果；然后，第一组给予B 处理，第二组给予A处理，观察结果。,常用实验设计方法,组间比较设计：两组或多组间进行比较 配对设计：直接配对分层配对 配伍组设计 拉丁方设计 析因设计 正交设计,实验误差及其控制,实验误差的必然性 系统误差： 仪器误差 方法误差 试剂误差 条件误差 顺序误差 人为误差,实验误差及其控制,随机误差 偶然误差 抽样误差过失误差,实验误差及其控制,实验误差的
11、控制 设立对照，严格按随机化原则抽样和分组 合理确定样本大小 正确选择实验方法与条件 杜绝过失误差,调查设计,调查的类型主要有：居民健康状况调查 流行病学调查 病因学调查 营养与卫生学调查 临床中、远期疗效调查,调查设计内容,调查目的 发病率、营养状况、疾病与环境 调查对象 调查范围 观察单位 调查项目（具体内容） 调查表格 调查指标,调查方法,根据调查范围： 普查 抽样调查 典型调查,调查方法,根据调查形式： 横剖面调查 特定时段内发生的事件 前瞻性调查 预先设定影响因素暴露组与对照组，在一定时间内观察它们的反应。 回顾性调查,COHORT (FOLLOW-UP) STUDY,Cohort
12、(follow-up) studies are observational analytical studies that are designed to mimic the randomized clinical trial. A cohort study does not involve investigator manipulation and, thus is not categorized as experimental, but it does test a hypothesis of a causal relationship and, thus, is analytical i
13、n approach.,COHORT (FOLLOW-UP) STUDY,A cohort is a group of persons, followed over time, having a common characteristic or factor of interest. A cohort or group is assembled on the basis of factors thought to relate to the development of the end point under study.,(continued),This allows investigati
14、on of a hypothesis concerning the etiology of the outcome of interest. The study involves following the group or cohort forward in time to observe its experience. The outcome studied most commonly is a disease.,COHORT (FOLLOW-UP) STUDY,Uses A cohort design is useful to determine the frequency of a n
15、ewly developed disease or a health-related event and to assess the exposure-disease relationship.,(continued),A drawback is that sufficient time must elapse between assessing the exposure and detecting the outcome, causing a typical cohort study to be long.,COHORT (FOLLOW-UP) STUDY,Unless the diseas
16、e studied is extremely common in the population to be studied, most individuals in a cohort will not develop it. Therefore, a large number of subjects is required in order to compare incidence between exposure groups.,(continued),Features The subjects in a cohort are apparently free of the disease u
17、nder study and are selected on the basis of the presence or absence of a factor of interest, termed an exposure. Subjects are then followed forward in time to determine the disease.,COHORT (FOLLOW-UP) STUDY,Selection of Subjects The subjects making up the cohort must be at risk for developing the di
18、sease or outcome of interest but free of the disease at the start of the study.,(continued),Sample Size The size of cohort required for the study is related to the frequency that the end point of interest occurs. A low frequency requires a large sample size.,COHORT (FOLLOW-UP) STUDY,Assessing Exposu
19、re Status Exposure to the factor of interest is observed or measured for each subject at the start of the study.,(continued),Assessing End Point, Disease The end point should be defined to be an unambiguous and reliable definition. The method and type of follow-up should be identical for all subject
20、s.,CASE CONTROL STUDY,Case-control studies are observational analytical designs that investigate hypothesis of causal relationship. These designs are retrospective, historically oriented studies, also known as case-referent or case-comparison studies.,(continued),Uses A case control design is used t
21、o explore etiology by comparing the prevalence of the exposure to factors of interest in persons who have a disease with that of a group without disease. It is useful in studies of rare diseases or end points. (The use and methods of case-control design: Journal of Chronic Diseases, 1979, 32:1-144).
22、 Less expensive and less time than cohort studies.,CASE CONTROL STUDY,Features The case-control study design assesses exposure status after disease status is known and thus is retrospective. The comparison groups are formed on the basis of disease or outcome status, either with disease diagnosis (ca
23、ses) or without disease diagnosis (controls).,(continued),Subjects are then investigated for the current presence of or previous exposure to a factor or factors of interest. The prevalence of the factor is compared between cases and controls.,Selection of Cases The goal in selection of cases is to o
24、btain a sample that is representative of cases arising from a defined target population. This goal is difficult or impossible to attain in most circumstances.,CASE CONTROL STUDY,(continued),The ideal compromise is to select all incident (newly developed or detected cases arising in a defined populat
25、ion over some specified period. Selecting incident cases rather than existing (prevalent) cases is preferred because factors related to survival, and, thus, to selection for the study, may differ from causal factors but cannot be distinguished when prevalent cases are used.,CASE CONTROL STUDY,Select
26、ion of Controls Selection of an appropriate group depends largely upon the hypothesis. The goal in selecting the control subjects is that controls should be representative of the population from which the cases arose.,(continued),Assessing Exposure The historical information, including the exposure
27、to the factor of interest, is obtained from existing records, examination, direct measurements, and personal interview or questionnaire.,The Ethics of Conducting and Presenting Research,At each step of research, ethical issues arise. Designing, conducting, reporting, and interpreting research all in
28、volve decisions in which professional ethics are pivotal.,(continued),Whenever human subjects are involved in research, the bioethics of the research must be recognized and carefully considered. Issues related to authorship and to conflict of interests are also of critical import in research ethics.
29、,Research error, human error, fraud,Research error, human error, and fraud Whether they are consequences of flawed design, improper conduct of research, or unintentional or intentional human error, scientific errors can seriously deflect scientific progress. Repercussion are many fold.,(continued),T
30、ime and finances can be lost in pursuing blind alleys, misapplication can be damaging to society, and scientific careers can be severely thwarted. Tainted literature, like an ocean blackened by an oil spill, requires time to be cleansed.,(continued),Poor and inadequate supervision is not acceptable
31、in scientific enterprises. It is critical that researchers assume responsibility and enable investigation if misconduct is charged in any research projects in which they have participated.,Research Errors may be categorized into six types: sampling error, non-coverage error, non-response error, meas
32、urement error, error of data distortion, and overgeneralization, and errors of misrepresentation to human subjects, in authorship, and in conflict of interests.,(continued),Human errors are a different class of error. Three sources of human errors need to be differentiated: inadvertent behavior, neg
33、ligence, and intentional actions. As scientists are fallible, inadvertent errors can occur. An honest mistake is tolerable to the scientific community and the public if it is promptly and properly handled when uncovered. However, preventable mistakes attributable to carelessness or negligence are no
34、t tolerable to either science or society; sloppy science is a form of intentional error.,Fraud, or intentional deception, destroys science by eroding trust and integrity. The scientific method is built upon hypotheses and honest observation; deception is anathema to science.,Fraud comes in varying d
35、egrees, including concealing data not supportive of a hypothesis and presenting only supportive data, revising observed data to conform to a hypothesis, and blatant fabrication of data. Each is an intentional deception. Plagiarism is a further category of intentional fraud.,Ethics in Research Involv
36、ing Humans,Nuremberg Code Following World War II, 20 German doctors were tried in Nuremberg before an international tribunal for war crimes and crimes against humanity.,The resulting Nuremberg Code of 1947 established ten principles that must be followed in human experimentation to satisfy moral, et
37、hical and legal concepts. These principles, for the first time, established as essential informed voluntary consent of human subjects.,Declaration of Helsinki The second major international code of ethics was the Declaration of Helsinki adopted by the World Medical Association in 1964. The basic pri
38、nciples in the Declaration of Helsinki were extended by the 29th World Medical Assembly in Tokyo in 1975, and further revised in 1983.,The 12 basic principles delineated the concept of submitting experimental protocols to an independent committee for consideration, comment, and guidance. The Helsink
39、i Declaration also counseled researchers to exercise caution in conducting research that could affect the environment and to respect the welfare of animals used for research.,Belmont Report A third important document supporting the rights of human subjects is the 1978 Belmont Report issued by the Na
40、tional Commission for the Protection of Human Subjects of Biomedical and Behavioral Research.,The Belmont Report addresses ethical conduct of research involving human subjects and argues for balancing societys interests in protecting the rights of subjects with its interests in furthering knowledge
41、that can benefit society as a whole. To assess benefit and risk of any research protocol, three principles are evoked: respect for persons, beneficence, and justice.,The Belmont Report also established recommendations for the protection of special categories of human subjects, including the human fe
42、tus, children, prisoners, and people institutionalized as mentally infirm.,Ethics in Designing, Conducting, and Analyzing Research,Ethical scientific conduct includes accurate recording of data in such a way that they are readily available and understandable to current and future colleagues.,To ensu
43、re appropriate data accessibility, the data need to be recorded at the time they are generated both correctly and in the detail necessary for ready comprehension. Original data books needed to be retained and made available if requested.,Throughout the research process, careful attention needs to be
44、 paid to details of subjects selection, method choice, and execution. If critical details are disregarded, if sloppy science is allowed, ethical predicaments may develop. This becomes of great concern when results are generalized.,Four Potential Research Errors,Sampling Error Errors of sampling resu
45、lt from the differences between the study sample and the actual population.,Ensuring that each individual within the study population has the same likelihood of participating will minimize random sample error. Another key element in minimizing sampling error is ensuring that the sample size is appro
46、priate to the goals of the research.,Non-coverage Error Non-coverage error results from a sampling format that excludes some individuals within the study population from being selected as subjects. Non-coverage errors are, in general, systematic and difficulty overcome.,They are caused by bias in sp
47、ecifying or selecting the study sample. Bias generally occurs when samples of convenience are selected. When bias cannot be eliminated, it is particularly important to recognize it and declare it to those who are considering the results of the study.,Non response Error The third category of error is
48、 that of non response. In surveys, the response rate is the percentage of those who actually answer the survey queries.,A low response rate raises serious questions as to weather the observed data accurately represent the study population: whether the non-responders differ significantly from the res
49、ponders. In many studies, researchers strive to improve response rate by devising various strategies to motivate, remind, and cajole subjects to respond.,Faulty Measurement Error Measurement error is the fourth type of research error. Whereas the first three types of error result from non-observatio
50、n, measurement error is one of observation. As such, errors in conducting or executing the research are akin to measurement error.,Well-Crafted Design The well-crafted research design minimizes research errors. All people in the population one wishes to observe have equal probability of being subjects, avoiding non-coverage error.,
