15 April, 2014: Food for Thought with Mark de Rooij

Presenter dr. Mark van Rooij outlined how in every step of research different choices have to be made and has discussed how we could set up a predictive investigation. The meeting was chaired by prof. Michiel Westenberg.

Mark de Rooij: Predictive Psychometrics

Photographer: Gerard Fritz

Photographer: Gerard Fritz

For an enterprise to be characterized as scientific it must have as its purpose to explain and predict phenomena. In psychology, statistical models are almost exclusively used for explanation. Explanation is equivalent to prediction only in deterministic systems where constructs (e.g., neuroticism) are immediately measurable and the form of the relationships (e.g., linear) between constructs is known.

Psychology, however, is not deterministic and therefore a good explanatory theory will not necessarily render a good prediction.


Absence of prediction
Prediction is largely absent in psychology; all we do is postdiction or retrodiction. The absence of prediction is damaging because:

  1. we lose the ability to test the relevance of existing theories and to discover new causal mechanisms;

  2. we erroneously infer predictive power from explanatory power;

  3. it raises confusion regarding the choice of methods, and

  4. it creates a gap between research and practice.

Importance of prediction
Prediction is important for diagnosis, prognosis, and decision making. Predictive questions that arise in practice are:

  1. Will this prospective student with characteristics as given follow our curriculum successfully/be successful in our degree course ?

  2. Will this depressed patient with characteristics as known (gender, personal circumstances, personality) recover within ten months if given cognitive behaviour therapy?

Different choices
The philosophy, epistemology, terminology, and methodology of prediction is completely different from that of explanation. In every step of research different choices have to be made in either setting. In this presentation Mark de Rooij will outline these differences and discuss how we could set up a predictive investigation.

More information

Last Modified: 17-04-2014