Lecture
In psychology, quasi-experiments are schemes for organizing data collection that make it possible to test psychological hypotheses while drawing on the standards developed in constructing research and conclusions on the basis of the experimental method.
A reduction in the forms of control is assumed here.
In a broad sense, a quasi-experiment is a way of planning psychological research and organizing the collection of empirical data that incorporates certain elements of experimentation but not all the stages implied by the unified logic of the experimental method.
In a narrower sense, this term is used by those authors who seek to emphasize the specific character of the forms of organizing research when it, like an experiment, is aimed at testing a cause-and-effect hypothesis but cannot be called one owing to insufficient control over the experimental manipulation and extraneous factors.
Examples of quasi-experiments in the narrow sense: various groups of designs:
- Between-group designs with full control of the IV but without fulfilling the condition of randomization.
- Between-group designs with analogs of the IV (the post factum control design).
- Designs with a single group of participants, in which changes associated with the time factor (a temporal trend) are represented.
(The basic variable (BV) – it is acted upon by the IV. It is correlated with the concept of basic processes.
The basic extraneous variable (BEV) – in the context of control over the conclusion (post factum) in a quasi-experiment – is a variable of internal conditions that is confounded with the main basic process under study.)
The goals and limitations of conclusions in quasi-experiments are conditioned by the following points:
1. The desire to investigate complex molar causal relationships that cannot be reproduced under laboratory conditions leads to experiments being conducted under field conditions, where chance factors (confounds) are controlled less well.
2. To remove such a threat to the validity of the conclusion as the participant's knowledge of the very fact of the experiment, a disguised experiment (a blind trial) is conducted.
3. Participants may also display other tendencies in changing their behavior owing to the actualization of special kinds of motivation (e.g., the motivation to be evaluated).
4. There are relationships that cannot be manipulated (internal, subjective conditions).
In the first three cases, various quasi-experimental designs with a deficit of control prior to the experimental manipulation are usually applied. In addition to these designs, one can make use of other means of control – control by means of additional variation and statistical control. For the fourth case, a qualitatively different means of control is applied, namely: control through the choice of «when and on whom to take measurements» (post factum control). That is, the experimenter selects groups of people for measuring the DV, assuming the differences to be already established or existing regardless of the research design (sex, motivation, and so on).
Two main directions of reducing the forms of control when forming experimental and control groups can be distinguished:
the failure to fulfill the condition of randomization when selecting participants;
the treatment, as an analog of the IV, of the difference between the groups that was introduced as the basis for the groups' non-equivalence.
1. The strategy of selecting groups by a given characteristic.
The selection of pairs of participants – this makes it possible to regard the quasi-experimental study as similar to a correlational one. The characteristic chosen as the variate may here be formulated as an analog of the DV or as an analog of the IV (the reconstructed causally acting variable). The participants matched into a pair will differ on some characteristic. It is precisely this differentiating characteristic that will underlie the difference between the homogeneous groups in the quasi-experiment. Example: Bandura and Walters – the connection between adolescent aggression and rejection within the family. The adolescents in each pair differed on the aggression indicator.
2. The design with a non-equivalent control group.
In educational research, the design with a non-equivalent control group is widespread (one of the quasi-experimental designs with a reduction of control prior to organizing the manipulations).
If already-formed real collectives are used in the experiment, for example study groups, then the experimental and control conditions cannot be considered equated, since differences between the groups are possible that may «overlay» the regularity under study and give rise to incorrect interpretations.
3. A mixed design of controlling external and internal conditions.
A variable of internal conditions (individual differences among people) is confounded with a variable of external conditions:
for example, propensity for risk (internal) influenced the DV (the decision to place in custody), and it was confounded with variables of external conditions (the availability of free places, in which situation the decision was applied).
Non-equivalent groups differ not in the levels of the experimental manipulations but in a parameter specified by the experimenter. The functional control of the levels of the variable (an analog of the IV) replaces the organization of manipulations.
4. The application of statistical control.
Increasing the number of participants to detect confounds of the BV with the BEV.
Unlike in a correlational study, conclusions here are constructed differently.
One part of the quasi-experimental schemes is built on the basis of true (in Campbell's terms) experimental designs, but here there are limitations in controlling threats to validity (e.g., because of the non-equivalence of the groups).
Another part – those that include only the researcher's choice of when and on whom to take measurements – is characterized by the absence of experimental control.
Here statistical control is carried out – control after the main empirical part of the study has been completed (post factum control).
For this, it is necessary to additionally measure the BEV (the sources of confounds).
- Post factum control
The general scheme is a system for comparing the influence of BEVs, measured as individual differences, on the basic process under study.
Control over the conclusion includes three stages:
1) The stage of initially selecting groups of different composition, differing in the levels of the main BV;
2) An experimental comparison of the DV values in these groups, the search for an «effect» of the BV;
3) Control over the confounding of the identified main effect with basic extraneous variables.
- The temporal trend
The time-series scheme presupposes the possibility of repeated (prolonged) measurement of the DV both before and after the manipulation regarded as experimental.
The resulting curves reflect the dynamics of temporal changes – these are the temporal trends. In analyzing a temporal trend, time is an analog of the IV. If the experimental manipulation changes the level over a given time interval, then this is a disruption of the temporal trend.
Types of quasi-experimental designs (?):
1. Experimental designs for non-equivalent groups;
2. Discrete time-series designs. Types: 1. The time-series experimental design 2. The time-sample series design 3. The equivalent-manipulations series design 4. The design with a non-equivalent control group 5. Counterbalanced designs.
Comments