Let's go If external noise and light affect the experiment, the best way is to enter the soundproof room or darkroom.
Samples can be excluded. Both Hawthorne effect and experimenter effect will affect the experimental results. The best way is to
Double-blind experiment was used. From the point of view of control variables, exclusion
The method really works. However, the research results obtained by exclusion method lack the universality of reasoning. For example, such as
If we are worried that the contact between subjects will affect the experimental results, we adopt automatic presentation of stimuli and
If the experimental results are recorded automatically, then the results cannot be used for similar behaviors in people's daily life.
Give an explanation by analogy.
2. Constant method Constant method is to keep the extra variables unchanged during the experiment. If it is difficult to eliminate redundant variables, the constant method can be used. Different experimental fields
Different laboratories, different experimenters and different experimental time are all additional variables. Effective control methods are the same.
A laboratory, by the same experimenter, uses the same experiment in both the experimental group and the control group.
Procedures for conducting experiments. If the noise with varying intensity cannot be eliminated during the experiment, a noise generator can be used.
Generate constant noise to cover it. In addition to the above experimental conditions remain unchanged, the experimenter and the control group
Characteristics of the subjects (such as age, gender, self-strength, achievement, motivation, etc.). ) is also the result of the experiment.
The main source of students' confusion should also remain unchanged. Only in this way can the differences between the two groups be resolved.
Due to the influence of independent variables. Using constant method to control redundant variables also has disadvantages: (1) The experimental results cannot be extrapolated.
Additional variables at other levels. For example, if only male adults are used as subjects.
Test, the results can not be extended to female adults. (2) manipulated independent variables and extra variables that remain unchanged.
Interaction may occur. For example, if the subject is male and the experimenter is attractive female,
During the experiment, the experimenter may distract the attention of the subjects, which is an additional variable produced by interaction.
3. Pairing method Pairing method is to let the subjects in the experimental group and the control group
Methods with the same characteristics. When using the matching method, all subjects should be measured first and the experiment should be completed.
Homework has the characteristics of high correlation; Then according to the measurement results, the subjects in the experimental group and the control group were divided into two groups.
The matching features are equal. If you want to do the experiment of "the influence of practice on learning to shoot successfully", make a prediction first.
Shoot to see the results of the subjects, and then put two subjects with equal predicted results (equal number of hits)
Subjects were divided into experimental group and control group respectively, and were matched into two groups of subjects with the same conditions to participate in the experiment. such
Although this method is desirable in theory, it is difficult to implement in practice because if there is more than one feature,
(or factors) above, experimenters often feel that they care about one thing and lose another, even worse. For example, experimenters
At the same time, we should consider factors such as age, gender, early maturity and intelligence, and try our best to match them.
It is difficult to divide into two groups equally. Even if this difficulty can be solved, many subjects can't take part.
Plus this experiment. What's more, motivation, attitude and other factors that belong to intermediary variables are even more impossible.
Find a reliable matching basis. Therefore, the matching method is not commonly used in practice.
4. Randomization is to randomly group the subjects according to the probability theory.
Assigned to each treatment group. Subjects were selected from the determined population by drawing lots or random numbers.
Sample, because random sampling makes all members of the population have the same opportunity to be extracted, so there is a considerable
It is very likely that the sample will keep the same structure as the population. After random sampling, the randomly selected samples will
Subjects' samples were randomly divided into different treatments. For example, there are three treatment groups: experimental group and real group.
Experimental group 2 and control group. Assign a number to each treatment group, such as 0, 1, 2, first.
Sample a, and then assign it to sample b and sample C. If "2" appears on the random table, set sample group A as control.
Group production; If "0" is encountered again, set sample group B as experiment 1, and so on. In theory, with
Machine method is the best way to control extra variables, because according to probability theory, each group of subjects has
All conditions and opportunities are equal and will not lead to systematic deviation. It can not only overcome the matching method.
It can also control intermediary variables (such as motivation, emotion, fatigue, attention, etc. ) is difficult to observe.
The random method is not only suitable for the subjects, but also suitable for the arrangement of stimuli. For example, if you have permission.
The subjects were treated several times to eliminate a series of effects (that is, the influence of the previous treatment on the later treatment)
Ring), you can arrange the order of various treatments in a random way.
5. Balance method The balance method is realized by adopting some methods.
The comprehensive balance method makes the influence of additional variables cancel each other out, and achieves the purpose of controlling additional variables.
Method. The main function of this method is to control the sequence effect. If the subject
When the experimenter applies a series of different treatments in a fixed order, the subjects' reactions will follow the order.
If the two treatments are not related in nature, it will have the effect of fatigue. These two shadows
Rings can confuse experiments, so cancel them out. If there are only two treatment methods, A and B, it is the most common.
The method to counteract the sequential effect is ABBA arrangement, that is, the same group of subjects give A position first.
Management, and then give B treatment; Then, in turn, treat B first and then A. If you're right,
The subjects in group B received more than two treatments, and the Latin square experiment can be used to offset the sequence effect.
Latin square experiment.
6. Statistical control method All the above methods can be used for experimental design. These methods are systematic.
It is called experimental control. But sometimes, due to the limitation of conditions, the above
All kinds of methods can't be used. We know that there are factors that will affect the experimental results, but we can't rule them out in the experiment.
Divide and rule. In this case, covariance analysis (or * * * analysis of variables) is only used after the experiment.
(covariance analysis), analyze the factors that affect the results, in order to achieve additional
Variable control. This method of controlling additional variables with statistical techniques afterwards is called statistical control.
System (statistical control). For example, students in two classes compare the two kinds of teaching in the experiment.
When the method is good or bad, although the experimenter knows in advance that the intelligence of the students in the two classes is not equal, it is limited by conditions.
The intelligence factors can't be controlled before the experiment, which makes the intelligence level of the two classes equal. Obviously, intelligence is
Important factors affecting experimental results. After the experiment, the influence of intelligence factors was analyzed by covariance analysis.
After exclusion, we can compare the advantages and disadvantages of the two teaching methods. The statistical control method not only analyzes covariance,
Partial correlation and other methods can also be used.