Loneliness and Health
- PDF / 477,422 Bytes
- 38 Pages / 504.57 x 720 pts Page_size
- 18 Downloads / 175 Views
Laboratory Stress Protocol ▶ Trier Social Stress Test
Latent Variable Jamil A. Malik National Institute of Psychology, Quaid-i-Azam University/VU University Amsterdam, Islamabad, Pakistan
Synonyms Hidden variable; Hypothetical Hypothetical variable
construct;
Definition Several of the concepts of social sciences extracted from various theories cannot be observed directly, although these concepts might be of interest to social scientists and researchers. For example: • In health: treatment adherence/compliance and well-being • In psychology: stress, self-worth, and personality dimensions • In education: memory, spatial ability, and problem solving
• In sociology: optimism, hostility, socioeconomic status, and attitudes • In political science: leadership, conservatism, fundamentalism, political competence, etc. In such cases where a concept is unobservable directly yet researchers are interested in their empirical testing, researchers have to rely on assumptions. In other words, they implicitly or explicitly operationalize these concepts into formal measurement models (constructs). These constructs can then only be measured indirectly via observable indicators/behaviors. Scaling (i.e., collection of relevant information on a questionnaire) is the most common technique to collect data for an unobservable construct. What is relevant is based on the assumptions of the researcher and derived from operational definition of the construct. Various types of scaling techniques have been developed for deriving information on unobservable constructs of interest from observable indicators. An important family of scaling methods is formed by latent variables. A latent variable model is a possibly nonlinear, path analysis or graphical model. In addition to the manifest variables, the model includes one or more unobserved or latent variables representing the constructs of interest. Latent variable models are developed on two basic assumptions. The first assumption is that the responses on the indicators are the result of an individual’s position on the latent variable(s). Secondly, it is assumed that
M.D. Gellman & J.R. Turner (eds.), Encyclopedia of Behavioral Medicine, DOI 10.1007/978-1-4419-1005-9, # Springer Science+Business Media New York 2013
L
1146
Latent Variable
Latent Variable, Fig. 1 Latent variable
LV
the manifest variables have nothing in common after controlling for the latent variable(s), which is often referred to as the axiom of local independence. Figure 1 shows a latent variable model, i.e., an unobservable variable LV represented by an oval which is assumed to be the result of commonality measured by three manifest indicators I1, I2, and I3. The ovals e1, e2, and e3 are residual also known as error variances. The concept of latent variables is based on the classical test theory, which assumes that any measure is a function of two variables: (1) the true score and (2) the error variation. This assertion can be written as a formula: (X ¼ T + E) where X represents the observed score on the measure, T is
Data Loading...
