Explore chapters and articles related to this topic
Cross-Lagged Panel Models
Published in Jason T. Newsom, Longitudinal Structural Equation Modeling, 2015
Inclusion of covariates is one strategy for addressing the potentially biasing effects of external variables that can affect stability or cross-lagged estimates. Available data never contain all possible factors that may have confounding effects, however, so any model is likely to be subject to omitted variable bias. With three or more time points, it is possible to test for spurious associations in the cross-lagged panel model that are due to omitted variables. There are two general strategies. One is to model a static “phantom” variable that represents all stable unmeasured covariates. The other is specification of a common factor that accounts for observations at each time point.
Health Information Technology
Published in Kelly H. Zou, Lobna A. Salem, Amrit Ray, Real-World Evidence in a Patient-Centric Digital Era, 2023
Joseph P. Cook, Gabriel Jipa, Claudia Zavala, Lobna A. Salem
The biggest challenges in RWD remain data and process related biases. Some of them are part of the research methodology in other disciplines, where qualitative or quantitative research methods are used. The National Institute of Standards and Technology (NIST, 2021) has drafted a proposal for identifying and managing bias in AI. Here we describe a few potential biases due to the following reasons. Data and data acquisition (Saunders, Lewis and Thornhill, 2009) Historical biasRepresentation bias (definition and sample from a population)Population bias (dataset is not representative to population statistics)Sampling bias (non-random sampling)Data collection and measurement errorDisparate systems with different ontologies and taxonomies.Algorithm induced (Mehrabi et al., 2019) Evaluation bias (during model validation)Popularity bias (news, search results, behavioral associations)Omitted variable biasCorrelated variables bias (inter-item similarity)Chained models and propagation of errorInterpretation induced Measurement bias (how choose and measure a particular feature)Aggregation bias (how we generalize based on subgroup)Behavioral bias (e.g., emoji usage in social media)Cause-Effect bias (e.g., correlation versus causation; The Stork-and-Baby Trap, 2021)
Measuring LGBT Discrimination in a Buddhist Country
Published in Journal of Homosexuality, 2023
Krichkanok Srimuang, Piriya Pholphirul
However this study also has certain limitations from this analysis. First, since this survey data comes from self-reporting, which can often be unreliable as there are many factors that can distort a person’s self-perception, such as their attitude toward life, their mind-set, or their life experiences that are not included in survey. In addition, econometrics analysis cannot control many important variables such as the “income variable.” Thus, inability to include many factors may have created an “omitted-variable bias” that cause bias toward the results.13 Second, it is also possible that cause an “endogeneity bias.”14 For example, those who discriminate more tend to conceptualize their behavior as “helping others” because they are so discriminatory that they do not think others deserve kindness to begin with. On the other hand, people who devote their lives to helping others often feel that, nevertheless, they have not done enough for others. Therefore, it is suggested that in order to improve the econometrics estimation, controlling the independent variables as well as using the instrumental variable (IV) technique should be considered in future studies.
Economic Stress and Life Satisfaction in Retirement Among Korean Older Adults: The Roles of Different Types of Social Support
Published in Journal of Gerontological Social Work, 2022
Jihee Woo, Hyojin Choi, Rafael Engel
Although this study suggests the underlying mechanisms through which economic stress relates to life satisfaction, there are limitations which necessarily inform our interpretation of results. First, the cross-sectional design of our study may limit the ability to determine causation. Longitudinal studies are required to determine a temporal and causal relationship between economic stress, social support, and life satisfaction. Unfortunately, questions about economic stress, social support, and life satisfaction were only asked in 2014, precluding such a longitudinal analysis with this data set. Additionally, despite our efforts to include well-known confounders in the structural model, unobserved or unobservable confounders may still exist, thereby resulting in omitted variable bias. Given the significant, yet not substantial, mediating role of social support in the relationship between economic stress and life satisfaction, confounding variables (e.g., resilience and leisure activity) may impact this relationship (S. Kim & Feldman, 2000; Manning, 2013). Other factors, such as depression, may affect both economic stress and life satisfaction in older adults. Future research should continue to address these questions in tandem with changing Korean demographic and socioeconomic trends, and the model should be tested in other countries. Overall, the results should be interpreted with caution and future investigations should improve upon these limitations.
Comparative effectiveness of budesonide inhalation suspension and montelukast in children with mild asthma in Korea
Published in Journal of Asthma, 2020
Jina Shin, Seung-Jun Oh, Tanaz Petigara, Kaan Tunceli, Eduardo Urdaneta, Prakash Navaratnam, Howard S. Friedman, Sung Woo Park, Song Hee Hong
Bias may have been introduced into the estimation of PDC because of inconsistency in recording days of supply for BIS nebules. There may also be an omitted variable bias and residual confounding associated with this study, given that pertinent clinical findings such as symptoms, lab results, or spirometry assessment data were not available for analysis. Exacerbation and asthma control measures that were utilized in this study were based on algorithms of readily available data and, thus, may have had limited applicability since other “true” exacerbation events defined by lung function or symptoms (e.g., nocturnal awakenings, wheezing, shortness of breath, etc.) were not captured. Finally, there are high rates of reported concomitant use of “over the counter” traditional medicines in Korea that may have either positively or negatively impacted asthma control. These data were not captured in this data source.