Thứ Bảy, ngày 28 tháng 2 năm 2009

Thống kê và nghiên cứu khoa học

Là tìm lời giải đáp cho những câu hỏi/ vấn đề chưa có câu trả lời bằng cách thu thập, phân tích diễn giải các chứng cứ khoa học theo các nguyên tắc và phương pháp khoa học phù hợp.

Research is the cornerstone of any science, including both the hard sciences such as chemistry or physics and the social (or soft) sciences such as psychology, management, or education. It refers to the organized, structured, and purposeful attempt to gain knowledge about a suspected relationship.

Research is a process of steps used to collect and analyse information in order to increase our understanding of a topic of issue. At a general level, research consists of three steps:
1. Pose a question
2. Collect data to answer the question
3. Present an answer to the question
This should be a familiar process. You engage in solving problem everyday and you start with a question, collect some information, and then form an answer. Although there are a few more steps than these three, this is the overall framework for reasearch. When you examine an published study or conduct your own study, you will find these three parts as the core elements.
When researchers conduct a study, they proceed through a distinct set of steps. Years ago these steps were identified as the “scientific method” of inquiry (Kerlinger, 1972; Leedy & Ormrod, 2001). Using a “scientific method”, reseachers:
• Identify a problem that defines the goal of research
• Make a prediction that, if confirmed, resolves the problem
• Gather date relevant to this prediction
• Analyze and interpret the data to see if it supports the prediction and resolves the question that initiated the research.
Applied today, theses steps provide the foundation for educational research. Although not all studies include predictions, you engage in these steps whenever you undertake a research study. As shown in Figure 1.2, the process of research consists of six steps:
1. Identifying a research problem
2. Reviewing the literature
3. Specifying a purpose for research
4. Collecting data
5. Analyzing and interpreting the data

6. Reporting and evaluating research.

(From: pp 3-8, Chapter 1, Educational Research – Planning, Conducting, and Evaluating Quantitative and Qualitative Research, John W. Creswell, 2nd ed, Pearson 2005, 2002)


+ phương pháp định lượng: sử dụng các con số
+ phương pháp định tính: sử dụng các chứng cứ không phải là số (ngôn ngữ hình ảnh âm thanh)

Quantitative versus Qualitative Research Methods

There are essentially two forms of educational research: quantitative (or statistical) and qualitative. In the past, most research was quantitative in nature, fashioned after the highly successful hard-science research methods. Since the 1990s, educational researchers have embraced qualitative research with the recognition that research on the human mind is fundamentally different from research on physical systems. Qualitative methods are used for depth of knowledge and quantitative methods are used for breadth and generalizability (emphasis added by PA). We cannot emphasize this enough: the best research designs employ both methods.
Most readers of this article will be familiar and comfortable with the ideas behind quantitative research. Because of this, and because quantitative research methods are well documented (e.g., Hopkins, 1998 ), we will not attempt a discussion of statistical research methods here. However, many readers will be uncomfortable with qualitative research because it is so different from the formal training of a biologist. Nonetheless, it is a vital component of educational research and should not be overlooked. What follows is a brief discussion of the most useful and common qualitative research methods for education.

Qualitative Research

Statistical research is well suited for drawing generalizable and repeatable conclusions based on a large sample of students, but it lacks depth. For example, suppose you use a particular approach in your course that you believe will lead to a greater understanding of transcription. You design a test that measures student understanding and give it to numerous students taught by using your approach and a traditional approach. You find that the students taught under your new method significantly outperform their peers. You come to the conclusion that your approach is superior. If your test and research methods were well designed, your conclusion is valid. You know your method is superior, but you have no data that tell you why the method was so successful (emphasis added by PA).. You probably have hypotheses about why the method worked so well (you had some reason to try it in the first place), but you do not know for sure. To properly answer the question of why, you must engage in qualitative research.

Qualitative research is also enormously helpful when you are initially investigating a particular area and are in the process of determining what questions might be interesting to study. The data of qualitative research are incredibly rich. Although this richness is an asset, it also makes qualitative data difficult and time consuming to analyze.

The National Science Foundation (NSF) has developed a detailed introduction to qualitative research geared toward the scientist. This introduction, User-Friendly Handbook for Mixed Method Evaluations, is excellent and can be found online at

PA: Là các nguyên tắc và phương pháp (quy trình, thủ tục) thu thập và phân tích các dữ liệu số (numerical data) về các hiện tượng để tìm hiểu bản chất và quy luật của các hiện tượng đó.

A statistic is a numerical representation of information. Whenever we quantify or apply numbers to data in order to organize, summarize, or better understand the information, we are using statistical methods. These methods can range from somewhat simple computations such as determining the mean of a distribution to very complex computations such as determining factors or interaction effects within a complex data set.

The science of statistics deals with the collection, analysis, interpretation, and presentation of data. We see and use data in our everyday lives. To be able to use data correctly is essential to many professions and in your own best self-interest.
(Collaborative statistics)

Statistics consists of the principles and methods for
1. Designing studies
2. Collecting data
3. Presenting and analysing data
4. Interpreting the results
Statistics has been described as
1. Turning data into information
2. Data-based decision making
3. The technology of the "Scientific Method"

Surfstat.australia: an online text in introductory Statistics
intro1.html 02/21/2009 16:23:37

Trả lời dựa trên trang web của surfstats
- Tóm tắt và trình bày số liệu
- Tạo số liệu mới (thiết kế nghiên cứu định lượng)
- Suy luận thống kê


What Are the Major Differences Between Quantitative and Qualitative Techniques?
As shown in Exhibit 1, quantitative and qualitative measures are characterized by different techniques for data collection.
Exhibit 1. Common techniques
Quantitative Qualitative
Existing databases Observations
Focus groups

Aside from the most obvious distinction between numbers and words, the conventional wisdom among evaluators is that qualitative and quantitative methods have different strengths, weaknesses, and requirements that will affect evaluators’ decisions about which methodologies are best suited for their purposes. The issues to be considered can be classified as being primarily theoretical or practical.
Theoretical issues. Most often, these center on one of three topics:
• The value of the types of data;
• The relative scientific rigor of the data; or
• Basic, underlying philosophies of evaluation.

Value of the data. Quantitative and qualitative techniques provide a tradeoff between breadth and depth and between generalizability and targeting to specific (sometimes very limited) populations. For example, a sample survey of high school students who participated in a special science enrichment program (a quantitative technique) can yield representative and broadly generalizable information about the proportion of participants who plan to major in science when they get to college and how this proportion differs by gender. But at best, the survey can elicit only a few, often superficial reasons for this gender difference. On the other hand, separate focus groups (a qualitative technique) conducted with small groups of male and female students will provide many more clues about gender differences in the choice of science majors and the extent to which the special science program changed or reinforced attitudes. But this technique may be limited in the extent to which findings apply beyond the specific individuals included in the focus groups.
Scientific rigor. Data collected through quantitative methods are often believed to yield more objective and accurate information because they were collected using standardized methods, can be replicated, and, unlike qualitative data, can be analyzed using sophisticated statistical techniques. In line with these arguments, traditional wisdom has held that qualitative methods are most suitable for formative evaluations, whereas summative evaluations require "hard" (quantitative) measures to judge the ultimate value of the project.

This distinction is too simplistic. Both approaches may or may not satisfy the canons of scientific rigor. Quantitative researchers are becoming increasingly aware that some of their data may not be accurate and valid, because some survey respondents may not understand the meaning of questions to which they respond, and because people’s recall of even recent events is often faulty. On the other hand, qualitative researchers have developed better techniques for classifying and analyzing large bodies of descriptive data. It is also increasingly recognized that all data collection - quantitative and qualitative - operates within a cultural context and is affected to some extent by the perceptions and beliefs of investigators and data collectors.

Philosophical distinction. Some researchers and scholars differ about the respective merits of the two approaches largely because of different views about the nature of knowledge and how knowledge is best acquired. Many qualitative researchers argue that there is no objective social reality, and that all knowledge is "constructed" by observers who are the product of traditions, beliefs, and the social and political environment within which they operate. And while quantitative researchers no longer believe that their research methods yield absolute and objective truth, they continue to adhere to the scientific model and seek to develop increasingly sophisticated techniques and statistical tools to improve the measurement of social phenomena. The qualitative approach emphasizes the importance of understanding the context in which events and outcomes occur, whereas quantitative researchers seek to control the context by using random assignment and multivariate analyses. Similarly, qualitative researchers believe that the study of deviant cases provides important insights for the interpretation of findings; quantitative researchers tend to ignore the small number of deviant and extreme cases.

This distinction affects the nature of research designs. According to its most orthodox practitioners, qualitative research does not start with narrowly specified evaluation questions; instead, specific questions are formulated after open-ended field research has been completed (Lofland and Lofland, 1995). This approach may be difficult for program and project evaluators to adopt, since specific questions about the effectiveness of interventions being evaluated are usually expected to guide the evaluation. Some researchers have suggested that a distinction be made between Qualitative and qualitative work: Qualitative work (large Q) refers to methods that eschew prior evaluation questions and hypothesis testing, whereas qualitative work (small q) refers to open-ended data collection methods such as indepth interviews embedded in structured research (Kidder and Fine, 1987). The latter are more likely to meet EHR evaluators' needs.

Practical issues. On the practical level, there are four issues which can affect the choice of method:
• Credibility of findings;
• Staff skills;
• Costs; and
• Time constraints.
Credibility of findings. Evaluations are designed for various audiences, including funding agencies, policymakers in governmental and private agencies, project staff and clients, researchers in academic and applied settings, as well as various other "stakeholders" (individuals and organizations with a stake in the outcome of a project). Experienced evaluators know that they often deal with skeptical audiences or stakeholders who seek to discredit findings that are too critical or uncritical of a project's outcomes. For this reason, the evaluation methodology may be rejected as unsound or weak for a specific case.

The major stakeholders for EHR projects are policymakers within NSF and the federal government, state and local officials, and decisionmakers in the educational community where the project is located. In most cases, decisionmakers at the national level tend to favor quantitative information because these policymakers are accustomed to basing funding decisions on numbers and statistical indicators. On the other hand, many stakeholders in the educational community are often skeptical about statistics and "number crunching" and consider the richer data obtained through qualitative research to be more trustworthy and informative. A particular case in point is the use of traditional test results, a favorite outcome criterion for policymakers, school boards, and parents, but one that teachers and school administrators tend to discount as a poor tool for assessing true student learning.
Staff skills. Qualitative methods, including indepth interviewing, observations, and the use of focus groups, require good staff skills and considerable supervision to yield trustworthy data. Some quantitative research methods can be mastered easily with the help of simple training manuals; this is true of small-scale, self-administered questionnaires, where most questions can be answered by yes/no checkmarks or selecting numbers on a simple scale. Large-scale, complex surveys, however, usually require more skilled personnel to design the instruments and to manage data collection and analysis.

Costs. It is difficult to generalize about the relative costs of the two methods; much depends on the amount of information needed, quality standards followed for the data collection, and the number of cases required for reliability and validity. A short survey based on a small number of cases (25-50) and consisting of a few "easy" questions would be inexpensive, but it also would provide only limited data. Even cheaper would be substituting a focus group session for a subset of the 25-50 respondents; while this method might provide more "interesting" data, those data would be primarily useful for generating new hypotheses to be tested by more appropriate qualitative or quantitative methods. To obtain robust findings, the cost of data collection is bound to be high regardless of method.

Time constraints. Similarly, data complexity and quality affect the time needed for data collection and analysis. Although technological innovations have shortened the time needed to process quantitative data, a good survey requires considerable time to create and pretest questions and to obtain high response rates. However, qualitative methods may be even more time consuming because data collection and data analysis overlap, and the process encourages the exploration of new evaluation questions (see Chapter 4). If insufficient time is allowed for the evaluation, it may be necessary to curtail the amount of data to be collected or to cut short the analytic process, thereby limiting the value of the findings. For evaluations that operate under severe time constraints - for example, where budgetary decisions depend on the findings - the choice of the best method can present a serious dilemma.

In summary, the debate over the merits of qualitative versus quantitative methods is ongoing in the academic community, but when it comes to the choice of methods for conducting project evaluations, a pragmatic strategy has been gaining increased support. Respected practitioners have argued for integrating the two approaches building on their complementary strengths.1 Others have stressed the advantages of linking qualitative and quantitative methods when performing studies and evaluations, showing how the validity and usefulness of findings will benefit (Miles and Huberman, 1994)



Introduction to Role of Statistics in the Scientific Method
Statistics has a major role to play in all stages of the scientific method. This is because it is involved with the definition and evaluation of hypotheses through the collection and analysis of data. In the paths (a) --> (b) and (e) --> (b) of analytical and inductive reasoning, the methods of descriptive statistics have their role to play. They provide powerful tools for suggesting questions to ask and formulating hypotheses. This is particularly useful in the study of large data sets, especially those routinely collected without specific research purposes; in mind. Such data should also be examined for indications as to the hypothesis or theoretical model underlying the process which produced the data. Data examination may include exploratory techniques such as tabulations, summary descriptions, graphical analysis, and cluster analysis.
Statisticians play an invaluable role in this exploratory stage by working closely with researchers. A basic understanding of the subject area and excellent communication skills are important for the success of this collaboration.
The experimental process, paths (c) --> (d) --> (e), of the scientific method is intimately is involved with many areas of statistics. A description of the statistical methods and thought processes in this part of the scientific method is depicted in Figure 2 and will now be discussed.

After clearly formulating the statistical hypothesis, relevant and valid data are accumulated from historical records, sample surveys or experiments in order to test the given hypotheses and provide indications for possible alternatives. Statistics provides the researcher with an array of methodologies to help in the design of an efficient and cost-effective data collection scheme which also ensures the accuracy, unbiasedness, and quality of the data.

in the area of measurement process the statistician's technical skills are needed. Close collaboration with researchers in the subject area and communication of statistical principles are also crucial.

The principles of quality control may prove to be of valuable assistance during and after data collection. Data ought to be routinely checked for the presence of errors, biases and outliers. The relevance of the data to the hypotheses under study also needs to be continually checked.

Statistics plays a crucial role, in experimentation. Even in the best planned experiments, we cannot control all the factors that affect our observations and we can rarely make. measurements without some noise or error from the measurement process. Hence, we have to make inferences based on imprecise sample data (emphasis by PA). To be of practical use, these uncertain inferences must be accompanied by probability statements expressing the degree of confidence the researcher has in the conclusions. To make certain that such probability statements will be possible, the experiments should be designed in accordance with the principles of statistical experimental design. These principles, together with the statistical hypothesis under study, dictate a statistical model relating the data to the statistical hypothesis through probability theory.

In other words, data have no meaning in themselves; they are meaningful only in relation to a statistical model of the phenomenon being studied. The interpretation of a set of data would be different, depending on what model was thought appropriate. In practice, some basic knowledge of the phenomenon under study is usually available to allow the researcher to specify a plausible statistical model.


- Rút ra quy luật thông qua việc quan sát các hiện tượng lập đi lập lại, sử dụng phương pháp quy nạp và kinh nghiệm thực (thực nghiệm)
- Tư duy có hệ thống và khoa học dựa trên các quy luật của số lớn
- Cho phép suy đoán về tổng thể từ một mẫu nhỏ hơn

Không có nhận xét nào:

Đăng nhận xét