Explore chapters and articles related to this topic
Vocabulary, Concepts and Usages of Structural Equation Modeling
Published in Douglas D. Gunzler, Adam T. Perzynski, Adam C. Carle, Structural Equation Modeling for Health and Medicine, 2021
Douglas D. Gunzler, Adam T. Perzynski, Adam C. Carle
Directly observable phenomena, such as height or weight, are routinely captured by a single measure. Measurable (directly observed) variables are also called manifest variables. Researchers may not have a single observed measure that adequately describes phenomena such as depression, pain behavior, happiness or anxiety. Multiple measures are often used to obtain reliable estimates, particularly for self-report or rater administered survey questionnaires. Medical researchers and practitioners alike need reliable and valid measures in order to understand symptoms, conditions and other clinically relevant measures that cannot be quantified directly. Variables that are not directly observed are considered latent variables and are also regularly labeled latent traits and/or latent constructs. Latent variable analyses can be useful for assessing the measurement properties of clinical screening, assessment and symptom scales.
Validity, Invariant Measurement, and Rater-Mediated Assessments
Published in George Engelhard, Stefanie A. Wind, Invariant Measurement with Raters and Rating Scales, 2017
George Engelhard, Stefanie A. Wind
The first source of validity evidence described in the Standards for Educational and Psychological Testing is evidence based on test content. This form of evidence is based on “an analysis of the relationship between the content of a test and the construct it is intended to measure” (p. 14). From the perspective of invariant measurement, this source of validity evidence can be conceptualized in terms of the Wright map. As noted in Chapter One, Wright maps are visual representations of the latent variable that facilitate the interpretation of locations of individual persons, items, and other facets in terms of this underlying continuum. For selected-response items, the key underlying question related to test content is: What is the latent variable being measured? When constructing a measure, researchers must begin by conceptualizing the underlying continuum in terms of a qualitative order from low to high (i.e., fail/pass; strongly disagree/disagree/agree/strongly agree; etc.). Examination of the alignment between observed and expected ordering of persons, items, or other facets on the Wright map can be used to support the interpretation of logit-scale locations as representing locations on the latent variable based on observations collected during the measurement process.
Advances in social measurement: A Rasch measurement theory
Published in Francis Guillemin, Alain Leplège, Serge Briançon, Elisabeth Spitz, Joël Coste, Perceived Health and Adaptation in Chronic Disease, 2017
Measurement is distinguished by a unit. The idea of a unit can be illustrated readily with the idea of a beam balance for measuring mass. The two objects constructed or found to have the same mass can be demonstrated readily with a beam balance. The identical mass of the two objects can be declared the unit. Then, if the two objects with the unit mass are placed on one side of the balance, and a new object with a mass balances the beam, the new mass is two times the mass of the unit. The process can, in principle, be continued in establishing a measurement of mass of other objects in a relevant range for the balance. A second feature of measurement that the beam balance exemplifies is that to measure the mass of an object it is necessary to somehow manifest the mass. In the above example, the effect of gravity on the masses is manifested on the instrument, the beam balance, in a controlled way. It is common to suggest that social measurement requires manifestation of variables because, instead of being observable, they are latent. The above example indicates that manifesting a variable through some instrument is also necessary in the measurement of mass and is typical in the natural sciences. Therefore, the idea of a latent variable is a feature of all measurement and is not confined to social variables.
Legal involvement and substance use treatment engagement and outcomes
Published in Journal of Social Work Practice in the Addictions, 2023
Ann Cherie Carter, Cory B. Dennis
Structural equation modeling using multinomial logistic regression (significance set at <.05) was performed due to the nominal nature of the discharge outcome and the latent constructs engagement and legal involvement (Hosmer et al., 2013; Muthén, 2001). Stata was used for preliminary data cleaning and for estimating sample descriptive statistics. Mplus was used for the confirmatory factor analyses (CFA) and for the structural equation modeling. Legal involvement and engagement were modeled as latent variables and the corresponding CFAs of indicators were driven by theory of what individual client engagement and legal involvement would include (Schreiber et al., 2006). After the measurement models for the latent variables were examined, the indicators drug screens and labs were correlated for sufficient model fit and because they both measure clients participating in a medical test. We then analyzed the structural model that included the exogenous latent variables legal involvement and engagement, and discharge type as the outcome. We first ran this model with age as the only control variable in accordance with Kline’s (2016) suggestion of hving 10 cases for every parameter. We also tested an alternative model that includes additional control variables: treatment location, times in treatment, and sex. As this did not compromise the structural model, we selected it as the primary model from which to report our main results.
Emotion dysregulation, help-seeking attitudes, and posttraumatic stress disorder symptoms: A structural equation model
Published in Journal of American College Health, 2023
Anne R. Limowski, Christopher R. DeJesus, Erin F. Ward-Ciesielski, Michael J. McDermott
Structural equation modeling analyses were conducted in Mplus version 7.243 to examine the relationships between latent constructs of Emotion Dysregulation, Help Seeking, and PTSD. This study treated total scores from questionnaires as indicators for the measurement and analysis of latent variables. Total scores from the DERS, RRS, PSS, and ASI-3 served as observed indicators of a latent construct, which was labeled Emotion Dysregulation. Likewise, total scores from the ATSPPH-SF, SSRPH, and personal/emotional problems and suicidal thoughts subscales from the GHSQ were used as indicators for a latent variable labeled Help Seeking. SSRPH total scores were reverse coded to match directionality with other indicators to facilitate accurate model convergence (i.e., higher scores indicate less stigma toward seeking help). PTSD symptom severity was modeled as a single-indicator latent variable using PCL-5 total scores with a fixed variance and internal consistency reliability as estimated in the current sample.59
An approach for knowledge acquisition from a survey data by conducting Bayesian network modeling, adopting the robust coplot method
Published in Journal of Applied Statistics, 2022
Derya Ersel, Yasemin Kayhan Atılgan
According to the conducted surveys in social sciences, many characteristics, such as support of a political party, economic progress, racial discrimination, alienation, irregularity, domestic violence, or religious loyalty can not be directly observed. However, we can use some observable characteristics which are indicative of them. For example, even though religious loyalty is not directly observable, the frequency of going to worship and prayer can be measured directly. It is usually thought that the observed variables are affected by latent or unobserved variables. So, the latent variables are defined by taking the relations between observed variables into account. Several statistical methods have been proposed to analyze these relations and define latent variables. One of these methods is the Latent Class Analysis (LCA). In a general sense, LCA is used to identify multi-level discrete latent variables using two or more cross classified observed variables [35].