Generalized Linear Factor Score Regression: A Comparison of Four Methods

Andersson, Gustaf; Yang-Wallentin, Fan

doi:10.1177/0013164420975149

Cited by 10 publications

(11 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, it is necessary to consider explanatory variables that reflect this interaction when disease risk is estimated using human gut microbiome information. SEM is a useful method that can achieve this because it constructs latent variables consisting of indicators with common characteristics (Dragan and Topolšek, 2014;Andersson and Yang-Wallentin, 2021). Furthermore, the number of latent variables and combination of indicators of latent variables can be set freely; therefore, Receiver operating characteristic (ROC) curve for a logistic regression model using lv1 or lv2 as the explanatory variables.…”

Section: Discussionmentioning

confidence: 99%

“…Structural equation modeling contains path analysis ( Dragan and Topolšek, 2014 ; Andersson and Yang-Wallentin, 2021 ); therefore, the influence of latent variables on the objective variable can be estimated as path coefficients. In women, the path coefficient from lv1 to the atopic dermatitis incidence status showed a variance of 0.32 ± 0.11 ( p < 0.01), and the path coefficient from lv2 to the atopic dermatitis incidence status showed a variance of −0.41 ± 0.12 ( p < 0.01; Figure 3 ) in our SEM.…”

Section: Discussionmentioning

confidence: 99%

“…Next, a SEM was then constructed to estimate the values of lv1 and lv2 when the participant’s atopic dermatitis status was unknown ( Figure 3 black frame). In general, the value of each latent variable in a SEM is calculated using all observed variables and all indicators in the model ( Dragan and Topolšek, 2014 ; Andersson and Yang-Wallentin, 2021 ). However, when using gut microbiota analysis data to estimate the risk of disease, it is necessary to estimate latent variables in the SEM with the observed variables if disease prevalence is unknown.…”

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

Method for estimating disease risk from microbiome data using structural equation modeling

Tokuno¹,

Itoga²,

Kasuga³

et al. 2023

Front. Microbiol.

View full text Add to dashboard Cite

The relationship between the human gut microbiota and disease is of increasing scientific interest. Previous investigations have focused on the differences in intestinal bacterial abundance between control and affected groups to identify disease biomarkers. However, different types of intestinal bacteria may have interacting effects and thus be considered biomarker complexes for disease. To investigate this, we aimed to identify a new kind of biomarker for atopic dermatitis using structural equation modeling (SEM). The biomarkers identified were latent variables, which are complex and derived from the abundance data for bacterial marker candidates. Groups of females and males classified as healthy participants [normal control (NC) (female: 321 participants, male: 99 participants)], and patients afflicted with atopic dermatitis only [AS (female: 45 participants, male: 13 participants)], with atopic dermatitis and other diseases [AM (female: 75 participants, male: 34 participants)], and with other diseases but without atopic dermatitis [OD (female: 1,669 participants, male: 866 participants)] were used in this investigation. The candidate bacterial markers were identified by comparing the intestinal microbial community compositions between the NC and AS groups. In females, two latent variables (lv) were identified; for lv1, the associated components (bacterial genera) were Alistipes, Butyricimonas, and Coprobacter, while for lv2, the associated components were Agathobacter, Fusicatenibacter, and Streptococcus. There was a significant difference in the lv2 scores between the groups with atopic dermatitis (AS, AM) and those without (NC, OD), and the genera identified for lv2 are associated with the suppression of inflammatory responses in the body. A logistic regression model to estimate the probability of atopic dermatitis morbidity with lv2 as an explanatory variable had an area under the curve (AUC) score of 0.66 when assessed using receiver operating characteristic (ROC) analysis, and this was higher than that using other logistic regression models. The results indicate that the latent variables, especially lv2, could represent the effects of atopic dermatitis on the intestinal microbiome in females. The latent variables in the SEM could thus be utilized as a new type of biomarker. The advantages identified for the SEM are as follows: (1) it enables the extraction of more sophisticated information when compared with models focused on individual bacteria and (2) it can improve the accuracy of the latent variables used as biomarkers, as the SEM can be expanded.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Method for estimating disease risk from microbiome data using structural equation modeling

Tokuno¹,

Itoga²,

Kasuga³

et al. 2023

Front. Microbiol.

View full text Add to dashboard Cite

show abstract

“…The second stage consisted of estimating factor scores based on the proposed measurement models, which serve as indicators of the relative location of each individual on the corresponding latent trait [ 37 ]. Then, a generalized linear model from a binomial family and a logit link function were carried out [ 38 ], considering somatic symptom severity as a dichotomous dependent variable with levels none to low (0), and moderate to severe (1). Factor scores were introduced as independent variables along with other sociodemographic characteristics from individuals.…”

Section: Methodsmentioning

confidence: 99%

Prevalence and Predictors of Somatization in Peruvian Undergraduate Students during the COVID-19 Pandemic

Zegarra-López

Bona

Romero-Montenegro

et al. 2022

IJERPH

View full text Add to dashboard Cite

The COVID-19 pandemic had a strong impact on mental health. Multiple studies report the alarming prevalence of depression, anxiety, and stress-related conditions due to the lockdown measures. Nevertheless, somatization has been an overlooked topic in current literature despite its strong relationship with most mental health conditions. The aim of this study was to describe the prevalence of somatic symptoms and their associated factors in a sample of 3218 undergraduate students from Lima, Peru. A cross-sectional design was carried out. The prevalence of somatic symptoms was measured with the PHQ-15. As predictors of somatic symptom severity, we included psychopathological (depression, anxiety, and stress), psychological (perceived social support, resilience, satisfaction with life, and academic self-efficacy), and sociodemographic (e.g., age, sex, employment status, relationship status, daily hours of sleep) variables. A generalized linear model from a binomial family and a logit link function were applied based on a Factor Score Regression approach, with half of the sample presenting moderate-to-severe somatic symptoms. Anxiety was the strongest predictor of somatic symptom severity, followed by academic self-efficacy. Significant differences were found regarding sex, relationship status, daily hours of sleep and COVID-19 risk-related variables. In conclusion, interventions on reducing anxiety and promoting academic self-efficacy may have a stronger impact on somatic symptom severity and should focus on more vulnerable specific demographic groups such as females.

show abstract

“…In practical applications, estimated factor scores ( F ˆ ) are sometimes used as proxies for factor scores ( F t ) . For example, in factor score regression models (Andersson & Yang-Wallentin, 2020; Devlieger et al, 2016), estimated factor scores are used as stand-ins for the latent factors. A potential problem with this practice is that the estimated scores do not fit the common factor model and often exhibit different properties from those of the model-consistent factor scores (Steiger, 1979b).…”

Section: The Factor Analysis Modelmentioning

confidence: 99%

Breaking Our Silence on Factor Score Indeterminacy

Waller

2022

Journal of Educational and Behavioral Statistics

View full text Add to dashboard Cite

Although many textbooks on multivariate statistics discuss the common factor analysis model, few of these books mention the problem of factor score indeterminacy (FSI). Thus, many students and contemporary researchers are unaware of an important fact. Namely, for any common factor model with known (or estimated) model parameters, infinite sets of factor scores can be constructed to fit the model. Because all sets are mathematically exchangeable, factor scores are indeterminate. Our professional silence on this topic is difficult to explain given that FSI was first noted almost 100 years ago by E. B. Wilson, the 24th president (1929) of the American Statistical Association. To help disseminate Wilson’s insights, we demonstrate the underlying mathematics of FSI using the language of finite-dimensional vector spaces and well-known ideas of regression theory. We then illustrate the numerical implications of FSI by describing new and easily implemented methods for transforming factor scores into alternative sets of factor scores. An online supplement (and the fungible R library) includes R functions for illustrating FSI.

show abstract

Generalized Linear Factor Score Regression: A Comparison of Four Methods

Cited by 10 publications

References 35 publications

Method for estimating disease risk from microbiome data using structural equation modeling

Method for estimating disease risk from microbiome data using structural equation modeling

Prevalence and Predictors of Somatization in Peruvian Undergraduate Students during the COVID-19 Pandemic

Breaking Our Silence on Factor Score Indeterminacy

Contact Info

Product

Resources

About