Research design

from Wikipedia, the free encyclopedia

The research design (also study design , study plan , test plan or test arrangement ) is the basis of every scientific investigation in areas of work that have to do with test persons or other living subjects . It is therefore especially important in social sciences , psychology , biology and medicine . It describes how the empirical question should be examined and defines which indicators should be recorded when, how often, where and how on which objects ( population , sample ). The established research design is decisive for the informative value of the research results.

There are basically two research approaches:

  • correlative studies
The structure of the relationships between the relevant variables is sought and quantitatively and systematically analyzed
Builds upon knowledge of the relevant variables and examines the nature of the relationship between the presumed predictors ( independent variable ) and the corresponding criteria ( dependent variable ).

Similar considerations and planning are also necessary for some experiments in natural sciences, technology and other areas. However, the terms described below essentially relate to the social sciences.

Correlative Studies

Correlation study

Is used when a separation into dependent and independent variables is not possible or the causality is not clear.

Ex post facto designs

Ex-post-facto arrangements are used when neither the requirements for experimental nor for quasi-experimental investigations are met. Both independent and dependent variables are measured and confounding variables cannot be controlled . For this reason, ex post facto designs only allow correlative statements. The advantage is that a large amount of data can be collected - usually in a survey  - with little financial and personal effort . Appropriate selection processes make generalizations possible. Ex post facto arrangements are the most common type of investigation in the social sciences. They can be divided into longitudinal and cross-sectional studies. Depending on the question of the investigation, another form of investigation is possible.

Longitudinal studies

In a longitudinal study , the same empirical study (usually a survey) is carried out at several points in time and the results of the individual research waves are compared with one another. A distinction is made between trend studies (also: replicative surveys ), panel studies and cohort studies.

  • Panel investigation, panel research , a special form of longitudinal studies in which the same people are recorded over a period of time. Example: The Socio-Economic Panel (SOEP)
  • Trend studies differ from panel studies in that different people are interviewed in each survey wave. However, these are always selected according to the same criteria. Example: ALLBUS
  • Cohort study : Here people of certain age groups ( cohorts ) are recorded. In the social sciences, cohorts are population groups that are defined by a common, long-term starting event. The different people are checked at different times.

Cross-sectional studies

Main article: Cross-sectional study

Cross-section (empirical research) in which different people are examined at the same time. In empirical research, one speaks of a cross-section or a cross-sectional study or cross-sectional design when an empirical investigation (e.g. survey, content analysis) is carried out once.

Pre-experimental set-up

The one-time treatment of a group and its “effect” measurement is referred to as a pre-experimental test arrangement “one-shot-case-study”. However, this form of research design must be viewed critically, since pre-experimental characteristics of the dependent variables and other influencing variables (independent variables) remain uncontrolled and alternative explanations cannot be ruled out.

Experimental designs

It is only an experiment if

  1. a distinction between independent (uV) and dependent variables (aV) is possible,
  2. the uV always precedes the aV and
  3. the data of at least two test subject groups are compared.

Experimental research designs test a hypothesis by manipulating the independent variable in a targeted manner and by controlling the influence of interfering variables by keeping the test conditions constant, elimination, randomization or parallelization. There are two types of experiments: laboratory and field experiments . The advantage of laboratory experiments is that the test conditions can be controlled to a high degree, which ensures high internal validity . On the other hand, field experiments often have the advantage that they have a high external validity due to the natural environment in which they are carried out .

Experiments are mainly used in psychology and communication science, to a lesser extent in the other social sciences; however, they are becoming increasingly important in economics.

Quasi-experimental designs

In contrast to experiments, the test subjects are assigned to the experimental and control groups in studies with a quasi-experimental test plan not by randomization or parallelization , but on the basis of existing properties of the test objects , such as age, gender, smoker / non-smoker, membership in a group, etc. For example, cultivation research asked about attitudes towards television viewing; People with high TV viewing were assigned to the experimental group, and people with low TV viewing were assigned to the control group. In the so-called “natural experiment”, the assignment is based on natural, uncontrolled events such as the introduction of a new type of school or the spread of a new medium. Quasi-experimental investigations do not allow any conclusions to be drawn about causal relationships , since it cannot be determined whether the independent variable causes the dependent or vice versa and whether both events are confounded .

Comparison: randomized experiment and quasi-experiment

Randomized experiments are characterized in particular by a random (randomized) distribution of the test persons to the experimental and control groups. In quasi-experiments , already existing characteristics of the test persons (e.g. daily television consumption) determine whether they are counted in the experimental or control group. The test plan of real experiments is called experimental design , the test plan of quasi-experiments quasi-experimental design .

The possible combinations of the above designs differ according to the table below with regard to internal and external validity ( quality criterion ). Internal validity exists if the change in the dependent variable can be clearly traced back to the variation in the independent variable (no alternative explanation). External validity exists if the result in the sample can be generalized to other people, situations and times.

Internal and external validity for experiment and quasi-experiment
External validity
External validity - also general validity , generalizability or ecological validity (see ecological fallacy ) - denotes the correspondence between the actual and the intended object of investigation. The basic idea here is the question of generalizability ( induction ).
One regularly conducts studies first on small and easily accessible entities, such as one's students or patients. A wrong generalization means e.g. B .: Although one should be warned of serious errors by many examples, it still happens very quickly and easily that a general validity is claimed for the results obtained in this way, which is often illusory.
For example, doctors often overestimate the severity and frequency of illnesses and complications because they only see those cases; Psychiatrists just as regularly underestimate the impact of psychiatric hospitalization and comorbidities because they have become used to them.
The correct procedure is to conduct a representative study after such an exploratory study ; Of course, this is time-consuming and sometimes very difficult.
Sample bias describes the deviation of a specific sample from the ideal of a strictly random selection from the correct population.
Internal validity
An experiment has a high internal validity (or ceteris paribus distributionibus validity ) if changes in the behavior of the test subject (dependent variable) are clearly due to the conscious change in the independent variable (treatment). In order to ensure this , interfering variables must be checked or switched off using various methods such as elimination, randomization, keeping constant and parallelization.
randomized quasi-experimental
field internal validity high / external validity high internal validity low / external validity high
laboratory internal validity high / external validity low internal validity low / external validity low

Field experiment

Laboratory experiment


  • H. Schnell, PB Hill, E. Esser: Methods of empirical social research . Oldenbourg, Munich 2005, ISBN 3-486-57684-4 , pp. 211-263.
  • ML Mitchell, JM Jolley: Research Design Explained . 4th ed. Clarion University of Pennsylvania, 2001
  • W. Hager: Basics of a test planning for testing empirical hypotheses in psychology . In G. Lüer (Ed.): General Experimental Psychology (43–253). Gustav Fischer Verlag, Stuttgart 1987.
  • DT Campbell, JC Stanley: Experimental and quasi-experimental designs for research . Rand McNally, Chicago 1966.
  • FN Kerlinger: Foundations of behavioral research . 2nd ed. Holt, Rinehart & Winston, London 1979.
  • G. Nieding, P. Ohler: Laboratory Experimental Methods . In: R. Mangold, P. Vorderer, G. Bente (eds.): Textbook of media psychology (Chapter 15). Hogrefe, Göttingen 2004.

Individual evidence

  1. H.-P. Musahl, C. Schwennen: Design of experiments . In: Lexicon of the editors: Gerd Wenninger. Spectrum Akad. Verl., Heidelberg 2000
  2. H.-P. Musahl, C. Schwennen: Design of experiments . 2000, p. 2
  3. W. Hager: Basics of a test planning for testing empirical hypotheses in psychology . In: G. Lüer (Ed.): General Experimental Psychology (43-253). Gustav Fischer Verlag, Stuttgart 1987.