1, comparison of three methods
2. Observation method
This is an original psychological research method, which is divided into natural observation, participatory observation, case study and investigation.
Natural observation means that researchers observe phenomena in natural scenes in order to obtain the true and natural behavior of subjects. For example, study the psychological impact of disasters such as earthquakes on victims.
Participatory observation means that researchers integrate themselves into the observation object group for observation. This makes up for the limited ability of natural observation to collect some information.
Case method refers to a long-term and comprehensive follow-up study of a single or several subjects by researchers. For example, luria's case study of people with magical memories.
Investigation method is an observation method that collects data with the help of questionnaires and scales, and makes analysis and processing to draw conclusions.
3. Related methods
Correlation method is used to observe and measure the statistical correlation between two or more variables, and determine their correlation strength and direction.
This method provides researchers with a correlation coefficient, which can evaluate and explain in detail the degree of correlation between two or more variables, expressed by R, and the value range is-1 to 1. Absolute value indicates the degree of correlation, symbol indicates the direction, 0 indicates irrelevance,-1 indicates negative correlation, and 1 indicates positive correlation. Commonly used correlation coefficients are Pearson product moment correlation coefficient and Spearman rank correlation coefficient.
In order to make the research results closer to the explanation of causality, the researchers put forward the cross-lag method.
The two correlation coefficients on the diagonal in the figure are cross-lag correlation coefficients, but this still cannot determine the causal relationship.
4. Experimental method
The experimental method has three basic elements, namely, experimental hypothesis and inference, experimental variables and experimental control.
The first element is experimental hypothesis and inference. The experimental hypothesis should follow the principle of operational definition and be transformed into operational reasoning.
The second element is experimental variables, including independent variables, dependent variables and additional variables. Independent variables are manipulated and controlled by the examiner, which have an influence on the response of the subjects. Dependent variable is the response of the subjects with the change of independent variables.
These two variables should follow the principle of operation definition.
Additional variables are potential factors and conditions that have nothing to do with the experimental purpose, but affect the dependent variables. They will confuse the influence of independent variables, so they need to be excluded.
The difference between experimental method and observation method and correlation method is that researchers are required to strictly manipulate independent variables and minimize the confusion of additional variables, so as to objectively observe the response of dependent variables to independent variables.
The third element is experimental control, which generally refers to all the work to ensure the accuracy of the experiment, and the most important thing is the control of additional variables.
According to different variables, experimental design can be divided into independent variable design, dependent variable design and additional variable design; According to the perspective of subjects, experimental design can be divided into inter-subject design, intra-subject design and mixed design; According to the number of samples, experimental design can be divided into large sample design and small sample design.
4. Experimental design of1variable
4. 1. 1, multi-independent variable design
A typical psychological experiment is a multivariable design that manipulates 2 ~ 4 independent variables at the same time. Its advantages are high efficiency, good experimental controllability and versatility, and it can reveal the interaction between variables.
As shown in the figure, in the free recall test, the scores of the two groups with different memory abilities are obviously different, but there is no obvious difference in the vocabulary recognition test, as shown in the figure below. Wallington and Wes Crantz opened up a new research field of implicit memory based on the discovery of interaction between independent variables.
4. 1.2, multivariate design
Multi-dependent variable design can obtain information that single-dependent variable experiment can't provide, but it is usually enough to use two dependent variables, because the dependent variables are used too much, and multivariate variance analysis is involved in statistical analysis, which is difficult and heavy workload.
Subjective and objective indicators are used to measure dependent variables.
Subjective index refers to the description of the subjects' psychological activities or the record of the answers to the examiner's questions.
There are six objective indexes: reaction speed, reaction speed difference, reaction accuracy, reaction standard, reaction difficulty and neurophysiological index.
Reaction speed indicators, such as measuring the number of tasks completed in a certain period of time;
The difference of reaction speed, such as the difference of reaction time between random letter group and regular letter group in implicit learning research;
Correctness indicators of response, such as measuring the number of times a maze has reached a dead end;
Standard indicators of response, such as measurement likelihood ratio? ;
The difficulty index of the answer, such as the three difficulty levels set by Skinner box;
Neurophysiological indicators, such as the subject's heart rate.
The effectiveness, objectivity and quantification of the dependent variable should be considered when selecting the indicators of the dependent variable.
First, effectiveness.
In other words, the index can fully represent the phenomenon studied. Among them, we should pay attention to avoid ceiling effect and floor effect. For example, if the maximum scale of the scale is lower than the weight of the measured object, the ceiling effect will appear. If the test questions are too difficult, resulting in high and low levels can not be done, there will be a floor effect;
Second, objectivity.
In other words, the index is objective and can be measured or verified by repeated experiments under certain conditions;
Third, quantification.
It means that indicators can be quantified.
In addition, the examiner should also guide the subjects to react within the dimensions envisaged by the examiner, because without guidance, the subjects' reactions will have no direction. At this time, the examiner needs to control and standardize the instruction, because the instruction has also become an important independent variable and should remain unchanged.
4. 1.3, extra variable design
Extra variables are divided into random extra variables and system extra variables. Errors caused by random extra variables are random errors, such as accidental temperature changes, which can be eliminated by multivariate statistical techniques. Errors caused by extra variables in the system are system errors, such as subjective effects and subjective effects, which can be controlled by the following six design methods.
First, the exclusion method
As the name implies, through a soundproof room, try to eliminate redundant variables, such as eliminating external noise.
If we want to eliminate the subjective experiment effect, we can eliminate it by single blind experiment. Subjective effect refers to two situations: the subjective will intentionally or unintentionally affects the subject with actions, expressions, languages, etc. In the experiment, make their reactions meet their expectations, or subjectively record data with biased assumptions when recording. Single-blind experiment requires that researchers who record dependent variable data do not know who is the experimental group and who is the control group, thus excluding the above two interference factors.
If we want to eliminate the subject effect and the subject effect at the same time, we can eliminate it through double-blind experiments. Subjects' effect means that the subjects will spontaneously make certain assumptions and guesses about the purpose of the experiment, and then respond in a way that they think they can achieve the purpose, or the subjects may intentionally feed back the data needed by the researchers, or they may agree with this assumption and choose to forget the non-hypothetical data intentionally or unintentionally, thus producing false data. Double-blind experiments require researchers and subjects to measure dependent variables, and neither of them knows who is the experimental group and who is the control group, thus eliminating the interference of the above factors. In addition, in order to prevent the subjects from guessing the purpose of the experiment, more control groups can be set up to prevent the subjects from guessing the hypothesis of the researchers easily.
Second, constant method
In practice, researchers can't exclude all the extra variables, so they can keep the extra variables unchanged during the experiment, which is equivalent to controlling the influence of the extra variables. For example, in order to overcome the deviation of experimental results caused by gender differences, researchers can choose a gender to conduct experiments.
But the limitation of this method is that it is difficult to generalize the experimental results to other levels of additional variables. For example, in order to control the influence of gender, the researchers only selected male subjects, so the conclusion cannot be directly extended to female subjects.
Third, the matching method
Matching method is to divide the subjects into several groups with the same characteristics as far as possible. For example, in the experiment of "the influence of practice on shooting effect", the researcher first tests the shooting performance of each subject, and then assigns the subjects with the same performance to match the two groups of subjects with the same shooting level.
This method is desirable in theory, but difficult in practice for four reasons.
First, the matching is often incomplete. It is impossible for researchers to match every feature of the subjects;
Second, matching is often time-consuming and laborious;
Thirdly, there may be interaction between matched features, which may confuse the experimental results. If the matched subjects are divided into junior high school students and college students, we want to study their memorizing ability of different materials, but in fact, age and memorizing materials interact, because the younger subjects have strong recognition ability of graphics and the older subjects have strong recognition ability of words, so if the memorizing ability of the two groups is different, it is not clear whether it is a material problem or an age problem;
Fourthly, there may be statistical regression illusion when matching. Statistical regression illusion means that the performance of extreme individuals tends to return to the average performance.
Fourth, randomization method.
Randomization means that subjects are randomly assigned to each treatment group to receive different independent variables. The logic is that if all members of the group have equal opportunities to be drawn into any treatment group, then the conditions and opportunities of each treatment group can be expected to be equal.
There are two specific methods of randomization: simultaneous distribution method and secondary distribution method.
The condition of simultaneous distribution method is that the subjects wait at the same time, and the examiner can dispatch any one of them at will. There are three skills in this operation: drawing lots, paddling and reporting. Among them, strokes refer to sorting by surname strokes first, and then assigning subjects with random number table.
The use condition of the second distribution method is that the subjects come at different times. There are two techniques for this operation: simple method and random method in the region.
Theoretically, the number of subjects randomly sampled by randomization method is usually more than 30, which is the best way to control additional variables. However, its limitation lies in that candidates can't accurately control the differences between different groups, that is, they don't know which variables are controlled.
Fifth, the offset balance method.
The offset balance method adopts ABBA method and Latin square design method to systematically change the presentation order of experimental treatment to offset or balance the errors caused by additional variables such as sequence effect, spatial effect, habit error, fatigue effect and practice effect. This method is one of the most commonly used methods in psychological experiments. Its logic is that if the subjects are arranged to receive all the sequential experimental treatments, then the difference in the results can be attributed to the independent variables rather than the order.
The operation of ABBA method is to divide the level of independent variables into A and B, and then let all subjects receive four experimental treatments in the order of ABBA. Theoretically, this method can effectively balance the position error and continuation error caused by the change of linear system.
Latin square design method is mainly used when the level of independent variables exceeds two. Latin square is a two-dimensional matrix, with columns representing experimental treatment and rows representing subjects. The formulas for establishing the first row of Latin square are 1, 2, n, 3, n- 1, n-2, and so on. As shown below.
Sixth, statistical control methods.
The first five methods are all controlled in the experiment, and this method is controlled by some statistical techniques after the experiment. This is because of the limitation of conditions, researchers know the existence of some extra variables, but they can't control them in the experiment, so they have to try to exclude the influence of these extra variables through some statistical techniques after the experiment. The main techniques are covariance analysis and partial correlation.
4.2 Experimental Design of Subjects
4.2. 1, subject room design
The design of the laboratory requires each subject to receive only one experimental treatment.
The biggest problem of this design is that because different subjects accept different experimental treatments, it is difficult to simply think that the change of dependent variables is caused by the difference of independent variables, and this difference may also be caused by the difference between subjects, so it is difficult to distinguish.
The way to make up for this defect is to minimize the differences between two or more groups of subjects, and two techniques can be adopted, namely, matching method and randomization method, which have been introduced in detail earlier.
The advantage of the test room design is that because everyone only accepts one kind of experimental treatment, the practice effect and fatigue effect can be avoided. However, there are two limitations, except that there is no way to fundamentally rule out the confusing effect of individual differences on the experimental results, and another point is that too many subjects are used.
4.2.2. Internal Design of Subjects
In-subject design requires each subject to accept all levels of independent variables and measure the behavior changes of the same subject under different treatments. Compared with the design between subjects, this design has no additional variables of individual differences because it looks at the changes before and after the subjects, but there are also confusion of additional variables such as position effect, continuation effect and difference continuation effect.
Position effect means that the position of the sequence processed in the experiment affects the reaction of the subjects.
The continuation effect means that the experimental treatment in the previous stage will have an impact on the later treatment. If the subject wants to receive two kinds of experimental treatments, he should receive treatment A first, and then when he receives treatment B, his behavior level may be improved or he is tired of experimental treatment because of treatment A, that is to say, the subject is affected by practice effect and fatigue effect.
Differential continuation effect also means that the previous stage of treatment affects the later stage of treatment, but unlike the continuation effect, this effect depends on what treatment comes first.
For these mixed effects, researchers use the balanced design method to eliminate and control them. ABBA method and Latin square design are commonly used, and these two methods are also introduced in detail.
Compared with inter-disciplinary design, intra-disciplinary design also has the advantage of saving the number of disciplines.
4.2.3, mixed design
Mixed design requires that one independent variable be designed within the subjects and the other independent variable be designed among the subjects. This design is actually carrying out two experiments, combining the advantages of internal design and inter-subject design.
4.3 Small sample design
4.3. 1, Comparison between large sample design and small sample design
Typical experiments usually adopt large sample design, because there are many samples, which can reduce the influence of individual differences between subjects on dependent variables.
However, in some special cases, a small sample design should be adopted, such as determining the effect of a certain therapy, but there are only one or two patients. Another example is psychophysical experiment, which is little influenced by individual differences, but it is time-consuming and laborious to do, so small sample design can also be adopted.
Small sample design has its own advantages. First of all, it helps to carry out exploratory research. Secondly, in the field of clinical psychology and educational consultation, it is particularly effective to verify the effectiveness of corrective procedures and test theoretical assumptions.
4.3.2. Two representative small sample designs
ABA design
A represents the baseline state before treatment and B represents the state after treatment. This design can be used to verify the effectiveness of the correction program.
ABA is to test the behavior of the subjects in the A stage, then treat the subjects, make them in the B state, measure whether the behavior changes, and finally restore the subjects to the A stage.
If researchers only adopt AB design, we can't directly attribute the change of B phase to the effect of treatment when measuring the change of B phase, because there may be other factors. Therefore, the key to ABA design lies in the last A. If the subjects can be restored to the A stage, the therapeutic effect can be distinguished, but the premise is that the therapeutic effect of the B stage cannot last, otherwise it will not return to the benchmark state of A.
Multi-baseline design
It refers to the introduction of the same independent variable in the experiment, but the introduction time is different, so that different behaviors or subjects will have different baseline periods before introducing the independent variable. The logic is that when one behavior is processed, the other behavior remains in the baseline condition. If the behavior at the baseline level remains stable before accepting the independent variable and changes after introducing the independent variable, we can conclude that the independent variable is the cause of the behavior change.
Multi-baseline design can be divided into multi-baseline design within subjects and multi-baseline design between subjects. Multi-baseline design within subjects is to compare the behavior of the same subject at different baseline stages, and multi-baseline design between subjects is to compare the same behavior of different subjects at different baseline stages.
5. Requirements
1. Distinguish the advantages and disadvantages of the three research methods.
2. What are the three basic elements of the experimental method?
3. What are the methods to control additional variables?
4. What are the connotations, advantages and limitations of different experimental designs?
Text/sailing against the current