Lecture
Correlational research refers to studies in which hypotheses about relationships are tested and in which the psychological variables are either manifestations of different aspects of the same basic processes, or accompany one another, while the question of their mutual determination remains open.
The main thing that distinguishes the correlational approach is its data-collection designs, which differ from the experimental ones.
In correlational research the variables are variates – variables whose changes can be measured but which cannot be manipulated (they are spontaneous in nature).
- One might speak of comorbidity (not sure why…)
- pathogenetic (cause-and-effect), diagnostic (one variable is caused by the action of two others and is their equally probable consequence), prognostic (two existing disorders predispose to the development of a third)
Correlational research designs should be regarded as forms of control in the acquisition of empirical data.
Example: Rushton’s correlational study. (personality characteristics of two groups of instructors: «researchers» and «teachers»). Examples from Gottsdanker:
- Ellen Siegelman, Jack and Jeanne Block, Anna Van der Lippe.
A comparison of groups of well-adjusted and poorly adjusted individuals in terms of the family conditions they had in childhood.
- Belmont, Marolla. A study of birth order and intelligence.
- A study for the selection of inspectors (questions of predictive validity).
Correlational designs, like experimental ones, also include a specification of the order in which data are obtained, but only as a plan for measuring the variables.
1. Designs with a single group of subjects. (QUESTION 26)
- each subject of the single group is represented in it by at least two measurements of variables, whether two different indicators or the same one assessed at a different point in time.
Hypotheses about relationships may in this case imply different ways of collecting data, since the data-collection plan depends on taking into account the substantive interpretations.
Variants of single-group designs:
1) Synchronic correlations
2) Autochthonous correlations
When using a single-group design, the researcher resorts to statistical control, which means controlling the variability of the variables (variates and EVs) by increasing the number of subjects in the sample.
2. Designs with two or more groups of subjects.
The design must also include forms of control of the EV. Three basic forms of control:
1) Stabilization of the EV, or the selection of levels of the values of the main variables so that the EV appears in the form of accounted-for levels that define a factorial design of its interrelations with at least one of the two measured variables.
2) Statistical control
- Design with homogeneous groups
A study of the dependence of IQ on birth order
3) Matching pairs; used in cases where the initial number of subjects is small.
The method of contrasting groups; a study of adjustment
3. Cross-lagged correlations as an approximation to causal inference. A shift over time in the changes of one of the variables relative to the other makes it possible to give preference, as to direction, to one of two competing hypotheses, leaving one of them. The two variables must be measured several times at the same time interval in the same subjects. Kenny’s study. The relationship between a child’s aggressiveness and the watching of programs with aggressive content (the variables were measured in children at the ages of 8 and 9 and at the ages of 18 and 19).
Comments