Experiment/quasi -experiment research in physics education research

Experiments/quasi -experiment research in Physics Education Research (referred to as PER) allow researchers to study the effects of some teaching improvements or intervention through comparison between two or more events. If this comparison involves quantitative analysis, inferring statistics are an effective tool [1]. In PER's experiment/quasi -experimental research, reasonable use of inference statistics cannot be separated from detailed research design [2].

1 Research and Physical Education Experimental Research and Physics Experimental Research

To put it simply, physical experimental research is to observe the change effect of another variable (due to variables) by changing the values ​​of certain variables (independent variables); and the study of physics education experiment also has similar situations. Researchers usually conduct events that they are interested in. Careful control and manipulation -introduce a difference in intervention, measure the differences brought by this intervention. In other words, educational researchers will also try to control something to determine the experimental effect. This kind of control is called an independent variable, and the independent variable can be one or more. By controlling the independent variables, the effect brought by it is due to the variable change. For example, in a "study of different teaching methods affecting college physical results", teaching methods are independent variables that educate researchers are interested in, and students' physical scores are due to variables.

The prerequisite for experimental research is to separate and control variables in order to establish and effect. Physics experiment research is usually performed in the laboratory, and the control and separation variables are easier to achieve from the theoretical perspective. It cannot be regarded as a research object that can be controlled at will, and there is no life. Therefore, the object of the research of physics education experiments is a complex system. It often faces uncontrollable internal and external interference and ethical dilemma. Researchers are basically impossible for researchers "True" experiments in the "ideal" environment usually adopt some experimental (quasi -experiments) design close to real (3].

2 Physical Education Experiment/Quasible Experiment Research Design

Let's start with the design of physics education experiment research to explain the essence of experimental research in the ideal state, and then introduce the design that can be used in practice in practice, so as to understand the research and design of physics education experimental research on how to seek optimal best between ideals and feasibility The plan is realized in feasible.

There are two terms in discussing the research and design of physics education experiments: pre -testing and post -testing, often appearing in studies related to collecting data. The previous test refers to the measurement or test of the subject before the experiment. The post -test refers to the measurement or test after the experiment. Not all experiments need to be performed in advance, and then the dependent variables for testing the experimental processing effect are necessary. The following briefly introduces two experimental research design and a quasi -experimental research design.

2.1 Experimental research design: only after the design of the control group design

The research and design of the physical education experiment generally includes two or more groups of students. One group (intervention) group (experimental group) in an experiment (intervention) may be included. There may be a control group. Accept experimental processing. At the end of the experiment, the research variables of the two groups were measured (post test). The design of the control group alone is an effective design for researchers, because it is a random group, and it does not require previous testing. The design of the control group only after the post -rear measurement control group is as follows:

Experimental group RG1 x O1

Control group RG2 --2

G means a group, R represents random extraction or grouping. X represents the experimental processing (independent variable), and the short -term indicates that there is no experimental processing. O represents the measurement of the variables, placed on the left/right side of X and short -line, which indicates the front/post -test respectively. In this design, there is only a rear test and no previous test.

Taking the "Study on the Fitness of College Physical Fitness" mentioned earlier as an example, using this research design, first randomly distribute students into the experimental group RG1 and the control group RG2. The experimental group uses a new teaching method to use a new teaching method Lecture course -experimental processing X (independent variable), the control group still uses the original teaching method -no intervention. After a semester of teaching, the physical scores (due to variable O) of the two class students can be tested and compared.

In the physics laboratory, heating a metal rod can be heated, and the researchers can affirm the expansion effect observed to the increase in temperature, because they can exclude (control) all other factors in the experiment. However, in PER experimental research, any variable variable may be affected by multiple factors. For example, the mutation of the physical academic performance of middle school students in the above studies may be affected by the teaching methods, which may also be affected by the ability, attitude, motivation, and previous previous. The effect of academic performance and other factors. The random distribution of students to the experimental group and the control group can theoretically reduce the effects of all possible unrelated variables on the variable. In other words, other unrelated variables except experimental variables (teaching methods) are similar to the effects (interference) in the experimental group and control group, and the "interference" effect produced Therefore, randomization ensures that the status of the experimental group and the control group is equivalent to as much as possible. In practice, the principle of randomization can only be used when there are enough trials. A good PER experiment research requires that researchers can provide a fair, non -deviation group comparison result. To this end, they must carefully ensure that the difference between any group can be attributed to the independent variables of research and design. 2.2 Experimental research design: Pre-test-post-test control group design

On the basis of "only post-test control group design", add previous test before experiment processing, which is "front test-rear test control group design", which is as follows:

Experimental group RG1 O1 x O2

Control group RG2 O3 — O4

Because it is a random group, the purpose of the previous test is not to ensure that the comparison group is equal. A pre -test of one or more variables can be statistically controlled during analysis and calculate the gain (GAIN), which is obtained by subtracting the previous test score.

In addition to using gain to indicate variables, researchers also often use standardized gains -the ratio between actual gain and potential maximum gain changes: actual gain: actual gain:

The actual gain is the difference between the latter score and the difference between the previous test score (GAIN = Post Score-Pre Score). The potential maximum gain is the difference between the full score and the previous measuring score. Since HAKE was introduced to Per for the first time [4], standardized gain has been widely adopted in physical education research. Researchers usually compare the standardized gain directly with 0 to check whether the performance of the students has changed significantly.

Per experimental research is characterized by random arrangements for trial (group) for experimental processing. However, in the research and practice of physics education, the selection and combination of the subject cannot always be random. Quasi -experimental studies refer to the use of original teaching classes in the experiment as a group for grouping, rather than randomly arranging the subject to receive it. If the researchers cannot randomly allocate the subject to enter each group, then the original group will vary in certain factors and cause each group to be different. It is risky using a group without random distribution, so researchers need to make efforts to get as close to the "ideal" experimental design as possible. This quasi -experimental design can make valuable contributions to education, but one thing is very important: researchers' interpretation and promotion of experimental results should be particularly careful. Here are a quasi -experimental research design.

2.3 Quasi -experiment research design: Pre -test -after -testing does not wait for the group design

In the previous "Experimental Research Design: Pre-test-Testing Control Group Design", if the experimental group and the control group are not randomly grouped, but using the original or naturally formed class to form the test group, it constitutes a "quasi-experimental design design The design of the front test — no -waiting group design "is as follows:

Experimental Group G1 O1 X O2

Control group G2 O3 -O4

The experimental group G1 and the control group G2 did not achieve peering through randomly. When we use the primitive or natural teaching class as the test group, we may bring sampling deviations. The similarity (or lack of similarity) between the two groups must be considered. To a certain extent Others outside the experimental intervention factors (independent variables) can affect the same as possible due to the interference of variables. The conclusions of the pre -testing are very helpful for the similarities between the testing experiment and the control group, because the previous test is a variable that is closely related to due to variables.

In the research of physics education, in addition to the above -mentioned experiment/quasi -experiment research design, there are also four groups of Romen design, factors design, repeated measurement design, time series design, etc. In addition, in order to apply the specific research situation, the researchers can make some changes to these experimental design. The important thing is that our design is suitable for experiments. Only by implementing full control experiments can the results be convincing and according to the results of the research results Promotion of applicability.

3 The characteristics of good physical education experiment research

In the study and design of the physical education quasi -experiment, in order to provide a fair, non -deviating group comparison data result, researchers need to try their best to ensure that the difference between any group can be attributed to the dependent variables of research. Do the following points: (1) reduce the confusion of variables or minimize this confusion; (2) distinguish and control irrelevant variables, reduce or reduce their impact; Test the conditions of the experiment and the accuracy of statistics. In theory, we use experimental validity to evaluate the quality of the experimental design. There are four types of experimental validity: the internal validity is concerned about whether the variables are caused by the variables; external validity refers to the experimental conclusion The exterior scope; structural validity is a clear definition of the structure of the independent variable and due to the variable structure; the validity of the statistical conclusion is to test whether there are differences and accuracy between the experimental group and the control group. Following the experimental requirements in physics experiment research. This requirement may not be particularly effective in physics education experiments/quasi -experiment research. Physics education experimental research is a variable that divides complex phenomena, and then only some variables that focus on it, just like the blindness of the blind people. The value of each research is that it provides some information, but it cannot obtain an overall perspective. The results of physics education experiments not only have sample dependencies, but also have differences between each individual in the sample. The ideal physical education experimental research is to randomly extract experimental samples from a certain overall sample, and then randomly distribute them into different groups. To. For example, a physical education experiment research sampled by a student of the liberal arts and physics courses of various universities conducted a sample. The results of the research results can be pushed to the liberal arts physics courses of different universities in terms of certain reliability, but you cannot lightly lightly. Putting out to the science and engineering physics course. When involving the validity of physics education quasi -experimental research, researchers must have a clear understanding of its defects, and determine the peers between the experimental group and the control group. At the same time, it is logically demonstrated by its possible representativeness and promotionability.

4 Data analysis of physics education experiments/quasi -experiment research

In the second article of this series of articles, we introduced the application of description statistics in the research of physics education [5], and physics education experiment/quasi -experiment research depends more on inference statistics for data analysis. The significant test of data differences is inferring statistics. The important content of the physical education experiment/quasi -experiment has a wide range of applications. The purpose is to judge whether there are significant differences between the two sets of data.

We still take the "study of different teaching methods to affect college physical results" as an example. The difference between the difference between the difference is not significant, indicating that the measurement variables of the two classes are equivalent at the beginning of the experiment. After a period of time teaching, the post -testing is performed. If the average test score of the experimental class is higher than the control class, and the results of the two sets of data difference test indicate that there is a significant difference between them. There are differences (such as student abilities, attitudes, motivations, differences in teachers, etc.), and the results indicate that the new teaching methods adopted by the experimental class are effective for improving student physical results.

There are many ways to infer statistics. There are detailed explanations in statistical textbooks. There are some methods that are complicated in calculations. Most data analysis can be performed through computers. Therefore Principles and applicable conditions. Researchers need to choose the appropriate inference statistical method based on research goals, sample types, data types [6], data distribution forms, etc., Table 1 gives some basic differential test methods and applicable conditions, and in the form of reference number Sign the application cases in PER, which is convenient for readers to consult.

Table 1 only provides several common statistical methods and applicable conditions in PER experiment/quasi -experiment research, not all. Among them, many professional terms are involved, and systematic learning is required to understand the statistics to understand.

Use a box of puzzles to compare it. This article tries to depict a simple picture printed outside the box. The purpose is to provide a guidance. The practitioner must open the box by himself, check and study each fragment, and strive to take them one by one to one by one to one by one Extract into a complete image. The field of physics education research is a diversified, complicated but very attractive research field [15]. It is also a very specialized and challenging field. Entering this field requires professionalism.

The community of physics education requires evidence as the basis for evaluation and action. The formulation of policy should be based on existing research. It is unreasonable to put an unrealized and inspection suggestion. The large -scale implementation is unreasonable. Physics education experiment/quasi -experimental research can enhance people's understanding of physics education, thereby improving physical education practice.

references

[1] FRAENKEL J R, Wallen N E and Hyun How to Design and Evaluate Research in Education [M]. New York: McGraw-Hill Higher Education, 2012. [2] Ding L, LIU X, Harper K. Getting Witch Quantitative Methods in Physics Education Research. Henderson C, Harper K.Eds.getting Started in Per [R]. Collegepark, md American association of physics 13, 2012, 3, 3,

[3] Ding L. Theoretical Perspectives of Quantitative Physics Education Research [J]. Physical Review Physics Education Research, 2019, 15 (2): 020101.

[4]HAKE R R. Interactive-engagement versus traditional methods: A six-thousand-student survey of mechanics test data for introductory physics courses[J]. American journal of Physics, 1998, 66(1): 64-74.

[5] DING L, Shen Yaqin, Zhang Ping. Research method of quantitative physics education: Investigation Research and Description Statistics [J]. Physics and Engineering, 2019, 29 (4): 3-6.

Ding L, ZHANG P, SHEN Y Q. Quantitative Methods in Physics Education Research: Survey Research and Descriptive Statistics [J]. Physics and Engineering, 2019, 29 (4): 3-6.

[6] Ding L, Zhang Ping, Jia Zehao. Quantitative research methods in physics education research [J]. Physics and engineering, 2019, 29 (2): 78-82.

Ding L, ZHANG P, JIA Z H. Quantitative Methods in Physics Education Research [J]. Physics and Engineering, 2019, 29 (2): 78-82. (In Chinese)

[7] POLLOCK S j. Longitudinal Study of Student Conceptual UndersTanding in Electrical and Magnetism [J].

[8]MCPADDEN D, BREWE E. Impact of the second semester university modeling instruction course on students' representation choices[J]. Physical Review Physics Education Research, 2017, 13(2): 020129.

[9] Stewart J, Griffin H, Stewart G. Context Sensitivity in the Force Concept Inventory [J]. PHYSICAL Review Special Topics-Physics Education Research, 2007, 3 (1): 010102.

[10] Day J, Bonn D. Development of the Concise Data Processing Assessment [J]. PHYSICAL Review Special Topics-PHYSICS Education Research, 2011, 7 (1): 010114.

[11]NIEMINEN P, SAVINAINEN A, VIIRI J. Force concept inventory-based multiple-choice test for investigating students' representational consistency[J]. Physical Review Special Topics-Physics Education Research, 2010, 6(2): 020109.

[12]BREWE E, SAWTELLE V, KRAMER L H, et al. Toward equity through participation in Modeling Instruction in introductory university physics[J]. Physical Review Special Topics-Physics Education Research, 2010, 6(1): 010106.

[13]TURPEN C, FINKELSTEIN N D. The construction of different classroom norms during Peer Instruction: Students perceive differences[J]. Physical Review Special Topics-Physics Education Research, 2010, 6(2): 020123.

[14] Clement J. Students' Preconceptions in intropuctory Mechanics [J]. American Journal of Physics, 1982, 50 (1): 66-71.

[15] Ding L, Zhang Ping. American physics education research: historical review and forward-looking [J]. Physics and engineering, 2018, 028 (1): 29-34.

Ding L, zhang P. Foundations and Frontiers of Physics Education Research in Thr U.S [J]. Physics and Engineering, 2018, 28 (1): 29-34. (In Chinese)

Fund Project: Hubei Province's provincial teaching research project in Hubei Province in 2017 "Research and Practical Research on Hybrid Learning Activities based on formation of formation" (2017276).

About the author: ding Lin, male, lifelong professor at the School of Education of Ohio State University, guest professor at Beijing Normal University, specially appointed professor of Chutian scholars in Changjiang University, research direction is physics education research and scientific education research, [email protected].

Communication author: Zhang Ping, female, professor, research direction is physics education research and scientific education research, [email protected].

Citation format: Ding Lin, Zhang Ping, Xu Yan. Experiment/quasi -experiment research in physics education research [J]. Physics and engineering, 2020, 30 (5): The first release of the network.

- END -