Latent Variable Regression Research Articles

Development and validation of a measure for father involvement during early childhood in a resource-limited context

BackgroundGlobally, there is mounting evidence about the importance of father involvement for improving child and family wellbeing. However, there are limited measurement tools available for assessing father involvement in low-resource settings globally. The objective of this study was to develop and validate a scale for measuring fathers’ engagement in caregiving activities for young children and families in rural Kenya.MethodsWe used baseline data collected in February–March 2023 from a cluster-randomized controlled trial of a parenting intervention aimed at improving early child development in Nyamira and Vihiga counties in Western Kenya. The analytic sample comprised 460 primary caregivers of children under 18 months of age (91.3% mothers) who were in a relationship with a male caregiver of the young child (i.e., father). The primary caregiver reported on the 25-item father involvement scale. We conducted exploratory and confirmatory factor analyses to determine the dimensionality of the measure and estimated latent variable regression models to assess construct and predictive validity in terms of associations with sociodemographic factors, other paternal and maternal psychosocial measures, and early child development outcomes.ResultsThe scale had a four-factor structure representing father involvement in (1) childcare activities, (2) play and affection, (3) household chores, and (4) early learning activities. Each factor showed good internal consistency reliability and evidence of multiple forms of validity. The four factors were associated with child and household sociodemographic characteristics and more positive paternal and maternal psychosocial outcomes. Father involvement across all four domains was associated with higher early child cognitive, language, motor, and socioemotional development scores.ConclusionsOur findings highlight the multidimensional nature of fathers’ caregiving roles in the lives of young children and their families. We establish a valid tool for assessing father involvement in rural Western Kenya that has potential for use in other similar contexts.

Soybean planting date affects the relationships between soil test values and grain yield

AbstractMany soybean [Glycine max (L.) Merr.] growers in the US Midwest rely on soil test values for evaluating the crop's fertilizer needs. However, threshold values for Illinois were calibrated to soybean yield in the 1960s when the production systems and yield potential were much different than today. The objective of this study was to determine which and how well soil test values predict yield of unfertilized soybean. Preplant soil samples, collected from 133 trials across Illinois from 2014 to 2021, were analyzed for 14 chemical attributes and compared to unfertilized soybean grain yields from those same studies. Pearson correlation coefficients (r), principal factor analysis, and latent variable regression models were used to determine those soil attributes most closely associated with grain yield and yield components. The association of planting date and yield (r = −0.56) led to dividing the data set into five planting date groups. Soil fertility levels resulted in a strong correlation with yield for Late or Very‐late planting groups, but not for the Early or Very‐early groups. A factor analysis of soil attributes largely resulted in retention of two factors, identified as Soil Organic Charge and Soil Fertility, across the planting‐date groups. Regression of these factors with yield confirmed that soil fertility had a greater influence on grain yield for late‐planted soybeans than early‐planted and that these differences were associated with average seed weight. Therefore, positioning late‐planted soybeans in higher fertility fields and early‐planted soybeans in lower‐fertility fields could reduce the need for supplemental fertilization.

Active Learning for Discrete Latent Variable Models.

Active learning seeks to reduce the amount of data required to fit the parameters of a model, thus forming an important class of techniques in modern machine learning. However, past work on active learning has largely overlooked latent variable models, which play a vital role in neuroscience, psychology, and a variety of other engineering and scientific disciplines. Here we address this gap by proposing a novel framework for maximum-mutual-information input selection for discrete latent variable regression models. We first apply our method to a class of models known as mixtures of linear regressions (MLR). While it is well known that active learning confers no advantage for linear-gaussian regression models, we use Fisher information to show analytically that active learning can nevertheless achieve large gains for mixtures of such models, and we validate this improvement using both simulations and real-world data. We then consider a powerful class of temporally structured latent variable models given by a hidden Markov model (HMM) with generalized linear model (GLM) observations, which has recently been used to identify discrete states from animal decision-making data. We show that our method substantially reduces the amount of data needed to fit GLM-HMMs and outperforms a variety of approximate methods based on variational and amortized inference. Infomax learning for latent variable models thus offers a powerful approach for characterizing temporally structured latent states, with a wide variety of applications in neuroscience and beyond.

Implications of confounding from unmodeled interactions between explanatory variables when using latent variable regression models to make inferences.

With linear dependency between the explanatory variables, partial least squares (PLS) regression is commonly used for regression analysis. If the response variable correlates to a high degree with the explanatory variables, a model with excellent predictive ability can usually be obtained. Ranking of variable importance is commonly used to interpret the model and sometimes this interpretation guides further experimentation. For instance, when analyzing natural product extracts for bioactivity, an underlying assumption is that the highest ranked compounds represent the best candidates for isolation and further testing. A problem with this approach is that in most cases the number of compounds is larger than the number of samples (and usually much larger) and that the concentrations of the compounds correlate. Furthermore, compounds may interact as synergists or as antagonists. If the modelling process does not account for this possibility, the interpretation can be thoroughly wrong since unmodelled variables that strongly influence the response will give rise to confounding of a first order PLS model and send the experimenter on a wrong track. We show the consequences of this by a practical example from natural product research. Furthermore, we show that by including the possibility of interactions between explanatory variables, visualization using a selectivity ratio plot may provide model interpretation that can be used to make inferences.

Examining the Conscientiousness × Interest Compensation (CONIC) model with similar constructs in high school and college students

The Conscientiousness × Interest Compensation (CONIC) model (Trautwein et al., 2019) assumes that the two constructs conscientiousness and interest can (partly) compensate for each other in predicting academic effort and achievement. We extend previous work by testing the CONIC model in two independent U.S. samples from different age groups (high school sample: N = 1,246, Mage = 14.86 years; college sample: N = 581, Mage = 19.83 years). We also assessed whether the compensatory relations occur when grit is substituted for conscientiousness and students’ overall task value or utility value for interest. We preregistered the study and tested the various models for both science and math effort as well as achievement, using latent variable regression analyses. Overall, we found some support for the CONIC model in the U.S. samples. There was also some support for the extensions of the model to the other constructs; however, the predictive links were not as strong.

Patterns of Polydrug Use Classes Among a Subset of Latino Male Seasonal Workers in South Florida: A Latent Class Analysis

ABSTRACT Background Little is known about polydrug use among Latino seasonal farmworkers. This cross-sectional study with male Latino seasonal workers (LSWs) living in South Florida categorized distinct classes of drug use and then characterized each drug use typology by demographic, structural, and psychological factors. Methods One hundred and twenty-five male LSWs were recruited during community events using convenience sampling between July 2019-March 2020. Latent class analysis was conducted by leveraging measures of self-reported use of nine drugs in the past year (sedatives, cannabis, stimulants, heroin, opioids, cocaine, PCP, hallucinogens, and inhalants). Correlates of latent class membership were examined using 3-step categorical latent variable logistic regression. Analyses were performed using Mplus version 8 and SAS 9.4. Results Four drug use classes were identified, male LSWs who exhibited: 1) use of illegal opioids only (n = 32); 2) concurrent cannabis and cocaine use (n = 75); 3) concurrent sedative and cannabis use (n = 13); and 4) high concurrent drug use (n = 5). About 84.7% of the sample reported use of at least one drug in the past 12 months. LSWs who identified as White and were married or in a stable relationship were likely to only use illegal opioids. LSWs with less than high school education and self-reported good to excellent health were more likely to use cannabis and cocaine. Men with moderate to severe anxiety and self-reported bad health were classified as engaging in sedative and cannabis co-use. Survivors of physical abuse were more commonly classified as people using high concurrent substances relative to non-abused men among the four classes. Conclusions This pilot study suggests the presence of heterogeneity in polydrug use classes among LSWs in South Florida.

Rain Pattern Deeply Reshaped Total Phosphorus Load Pattern in Watershed: A Case Study from Northern China

Excessive phosphorus in aquatic systems poses a threat to ecosystem stability and human health. Precipitation has a profound influence on the phosphorus biogeochemical process; however, it has been inadequately considered at the watershed scale. In this study, the Bayesian latent variable regression model was utilized to quantify the impact of rainfall on the concentration of total phosphorus using daily monitoring data from 2019 to 2021. The results revealed a piecewise linear relationship between total phosphorus concentration and precipitation. It was further inferred that the breakpoint (The total rainfall during a single rainfall event equal to 39.4 ± 0.45 mm) represented the tipping point for the transformation of the primary river runoff generation mechanism. Subsequently, the excess phosphorus load caused by rainfall events was estimated in the Shahe basin by combining the regional nutrient management approach with the results of the Bayesian latent variable regression model. The results indicated that rainfall events were one of the most significant sources of TP loads from 2006 to 2017, accounting for 28.2% of the total. Non-artificial land, including farmland, forests, and grasslands, serves as the primary source of the excess phosphorus load resulting from rainfall events. This study provides insights into how to quantify the phosphorus load resulting from rainfall events at the basin scale, which is valuable for phosphorus management. Environmental managers should prioritize the regulation of phosphorus in non-artificial land moving forward. Implementing hierarchical management based on calibrated curve numbers and soil phosphorus content could prove to be an efficient approach for regulating phosphorus in the watershed.

Are We Creatures of Logic or Emotions? Investigating the Role of Attitudes, Worldviews, Emotions, and Knowledge Gain From Environmental Interpretation on Behavioural Intentions of Park Visitors

Environmental interpretation can improve sustainability by mitigating the negative impacts of nature-based recreation. However, we do not fully understand the psychological factors that influence interpretation’s efficacy in changing human behaviours. Specifically, the role of emotions has been understudied within environmental psychology and nature-based recreation. This study, therefore, provides further insight into the psychological processes driving pro-environmental behavioural intentions among overnight visitors attending personal interpretation programs in provincial parks in Alberta, Canada. In 2018 and 2019, we surveyed 763 attendees of personal interpretation events. We used latent variable structural regression modeling to test the hypothesized relationships between ecological worldview, attitudes, emotions, and pro-environmental behaviours. As predicted, there were positive relationships between worldviews, affective and cognitive attitudes, and emotions; these variables and knowledge gain were positively associated with pro-environmental behaviours. Findings suggest that interpretation should focus programming on the affective elements of communication, target personal meaning such as responsibility to act, and continue to transmit knowledge.

An improved dynamic latent variable regression model for fault diagnosis and causal analysis

AbstractThe advancement of industrial techniques has imposed a high demand for powerful machine learning algorithms to model the increasingly complicated relations in the data. Among them, dynamic models are widely studied to capture the inevitable temporal relations. However, most existing methods only focus on the dynamics between input and output data, failing to exploit other valuable information in the output. In this article, an improved dynamic latent variable regression (LVR) method is proposed to capture both auto‐correlations and cross‐correlations between input and output with an auto‐regressive exogenous model, which is referred to as dynamic regularized LVR with auto‐regressive exogenous input (DrLVR‐ARX). Further, a DrLVR‐ARX‐based fault detection and diagnosis framework is designed to identify the root causes of a detected fault. The framework systematically integrates reconstruction‐based contribution, time‐domain Granger causality, and conditional spectral Granger causality to determine and locate the assignable causes. The effectiveness of the proposed algorithms is demonstrated with two industrial processes.

Improved Semi-Supervised Data-Mining-Based Schemes for Fault Detection in a Grid-Connected Photovoltaic System

Fault detection is a necessary component to perform ongoing monitoring of photovoltaic plants and helps in their safety, maintainability, and productivity with the desired performance. In this study, an innovative technique is introduced by amalgamating Latent Variable Regression (LVR) methods, namely Principal Component Regression (PCR) and Partial Least Square (PLS), and the Triple Exponentially Weighted Moving Average (TEWMA) statistical monitoring scheme. The TEWMA scheme is known for its sensitivity to uncovering changes of small magnitude. Nevertheless, TEWMA can only be utilized for monitoring single variables and ignoring the correlation among monitored variables. To alleviate this difficulty, the LVR methods (i.e., PCR and PLS) are used as residual generators. Then, the TEWMA is applied to the obtained residuals for fault detection purposes, where the detection threshold is computed via kernel density estimation to improve its performance and widen its applicability in practice. Real data with different fault scenarios from a 9.54 kW photovoltaic plant has been used to verify the efficiency of the proposed schemes. Results revealed the superior performance of the PLS-TEWMA chart compared to the PLS-TEWMA chart, particularly in detecting anomalies with small changes. Moreover, they have almost comparable performance for large anomalies.

Anticipatory transport system with hybrid linear and nonlinear forecasting using streaming wafer process data

In semiconductor plants, autonomous vehicle systems (AVSs) are designed to transfer wafer lots using several hundreds of vehicles. To minimize the idleness of production machines, vehicles must be quickly assigned and sent to the required location in advance. Currently, owing to uncertainties that exit in production processing, simple deterministic heuristic approaches are the ones that are predominantly used when making decisions regarding lot discharging and recharging of production machines. However, to obtain better solutions that reflect the near-future states of production machines and AVSs, more sophisticated approaches are required. To address this, we propose a hybrid predictive algorithm using which a delivery vehicle can be sent in advance to minimize the production machine idle time. To predict the remaining processing time, we combine a latent variable regression and an artificial neural network to model and generalize the linear and the nonlinear patterns of wafer takt times. We then use dynamic time warping to identify the best matching pattern for the wafers being processed, thereby accurately estimating the remaining processing time. The experimental results demonstrated the superior performance of the proposed method in comparison to the existing ones in terms of accuracy. Finally, based on the results of field applications, we found that the proposed method can improve the operational efficiency of AVSs by allowing fast wafer transfer and maximize production throughput by reducing machine idle time.

Variable Selection and Redundancy in Multivariate Regression Models

Variable selection is a topic of interest in many scientific communities. Within chemometrics, where the number of variables for multi-channel instruments like NIR spectroscopy and metabolomics in many situations is larger than the number of samples, the strategy has been to use latent variable regression methods to overcome the challenges with multiple linear regression. Thereby, there is no need to remove variables as such, as the low-rank models handle collinearity and redundancy. In most studies on variable selection, the main objective was to compare the prediction performance (RMSE or accuracy in classification) between various methods. Nevertheless, different methods with the same objective will, in most cases, give results that are not significantly different. In this study, we present three other main objectives: i) to eliminate variables that are not relevant; ii) to return a small subset of variables that has the same or better prediction performance as a model with all original variables; and iii) to investigate the consistency of these small subsets.

Kernel-Regularized Latent-Variable Regression Models for Dynamic Processes

Establishing relations between variables and real-time prediction of quality variables or other key indicators are critical for dynamic processes including industrial and biological processes. In this study, a novel multivariate statistical modeling method named “kernel-regularized latent variable regression (KrLVR) approach” is proposed for capturing the dynamics of a process by building KrLVR models with process and quality data. First, a regularization term based on a kernel matrix is incorporated into the objective of the latent variable regression model. Consequently, the proposed KrLVR method has the ability to overcome potential ill-conditioning resulting from collinearity in process data and has a stronger prediction power. Besides, the inner model is consistent with the outer model, which enables the proposed method to predict quality data with fewer latent variables. Second, the prior knowledge of dynamic processes such as exponential stability and smoothness can be integrated into the modeling process by using an appropriate kernel matrix. In addition, to meet the requirement of exponential stability of the model, the weights of the model should decay exponentially, that is, coincident with determining the number of historical observations for data augmentation (identification of the model structure). Therefore, the problem of tuning model complexity is eluded, and it becomes finding appropriate hyper-parameters of the kernel matrix. Moreover, the empirical Bayesian method is utilized for estimating hyper-parameters of the kernel matrix from augmented process and quality data. Three case studies illustrate the performance of the proposed KrLVR method by comparing with several other relevant methods.

Concurrent Multilayer Fault Monitoring with Nonlinear Latent Variable Regression

Latent variable regression (LVR) constructs the latent structure by maximizing the projection of quality variables on the latent spaces of the process variables, which has demonstrated its effectiveness over other multivariate statistical methods in terms of quality-relevant monitoring. However, it cannot be applied to model most real-world industrial processes due to its implicit linear assumption on the data. To overcome this issue, the kernel trick is designed to develop its nonlinear counterpart. Further, a novel concurrent kernel LVR (CKLVR) method is proposed for multilayer modeling with subsequent decomposition, which separates the process and data spaces comprehensively into five subspaces. The corresponding monitoring statistics are also developed for each subspace extracted by CKLVR, with each serving a particular monitoring purpose. Case studies including a numerical simulation, Dow’s Refining Process, and the Tennessee Eastman Process are employed to illustrate the performance of the proposed algorithm.

Land tenure security and farm efficiency in Indian agriculture: Revisiting an old debate

This paper examines the Marshallian inefficiency hypothesis of share tenancy system in agriculture of Odisha, Eastern India. It examines the hypothesis from a different perspective where it is argued that a secured land tenure system enhances farm efficiency even though it is sharecropping. We develop three hypotheses within the profit maximization framework and use the efficiency scores of paddy growers computed from the data envelop analysis (DEA) to verify them empirically. The ‘latent variable regression model’ results reveal that a secured land ownership status significantly augments farm efficiency. But the nature of tenancy contract (sharecropping and fixed-rent tenancy) makes no difference in farm efficiency. The policy implication of the study calls for tenancy reform in Odisha agriculture to make the tenancy a more secure one. Currently, the government of Odisha has banned the share tenancy. But a vast concealed tenancy comprising of marginal farmers and landless cultivators is in place. So any reform in this regard would benefit them directly.

Consumer willingness to pay for plant-based foods produced using microbial applications to replace synthetic chemical inputs.

Analysis of consumer preferences and willingness-to-pay (WTP) for sustainable foods produced using new agri-food technologies is required to enhance the uptake of innovations that accelerate the transition towards sustainable food systems. Consumers’ willingness to buy new food products, with no or limited consumption experience, mainly depends on their food choice motivational orientations (promotion- vs prevention-orientation). The objective of this study was to elicit consumers’ WTP for foods that are produced with microbial applications during the plant production phase with the aim to reduce the use of synthetic chemicals in crop farming, as well as to understand the associations of food choice motives, personal and socio-demographic factors with the WTP. We used contingent valuation to elicit consumers’ WTP for three food products (wheat bread, consumer potatoes and tomato sauce) through online surveys. Data were collected from 291 consumers, primarily from Italy, Germany and the Netherlands. Descriptive statistics, latent variable modelling and logistic regression were used to analysis data. Results show that more than two-third of the respondents are willing to pay premiums of at least 0.11 euro per kg of food products for reductions in synthetic chemical use by at least 50% due to microbial applications. The amount of WTP increases with the level of reductions in synthetic chemical use. The majority of the respondents are promotion-oriented consumers in relation to their food involvement, and are more likely to pay premiums for the sustainably produced food products. Environmentally concerned consumers are also more likely to pay premiums, whereas health concerned consumers are not. This study contributes to understanding of consumers’ attitude and perceived health risks towards foods obtained using microbial applications, and the heterogeneity of their preferences. Results provide insights for identifying potential buyers of foods produced using microbial applications, and to set prices according to the levels of consumers’ WTP.

Gaussian mixture deep dynamic latent variable model with application to soft sensing for multimode industrial processes

The data-driven deep probabilistic latent variable model (DPLVM) has attracted much attention for industrial process soft sensing in recent years. The DPLVM has well handled the nonlinear characteristics of the processes with powerful feature extracting capability. However, the multimode process property and the dynamic data features seldom be considered in those applications. To tackle the two issues, the article starts from the basic DPLVM, i.e., the Variational Autoencoder (VAE), to build a deep dynamic latent variable regression model (i.e., the Gated Recurrent Unit-based VAE regression, GVAER), where GRU cells are utilized to capture dynamic features from the process time sequence data. With the GVAER, a Gaussian Mixture GVAER (GM-GVAER) model is proposed. The Gaussian Mixture priors are used in the latent space to characterize the multimode process data features. In particular, a semi-supervised learning scheme is also proposed for the model to deal with the unequal scale of input and output data. A numerical example and a real chemical process case are provided to verify the feasibility and effectiveness of the proposed soft sensor model.

Tina's ParTy line: Polysubstance use patterns in sexual minority men living with HIV who use methamphetamine.

This cross-sectional study examined the associations of demographic, structural and psychological factors with distinct typologies of polysubstance use in sexual minority men (SMM) living with HIV who use methamphetamine. In total, 161 SMM living with HIV who reported methamphetamine use in the past 3 months were recruited in San Francisco from 2013 to 2017 for a randomised controlled trial. A latent class analysis was conducted by leveraging baseline measures of self-reported use of 15 substances in the past 3 months as well as validated screening measures of hazardous alcohol and cannabis use. Correlates of latent class membership were examined using a three-step categorical latent variable logistic regression. Four typologies of substance use were identified: (i) methamphetamine use only (43%); (ii) methamphetamine and crack-cocaine use (22%); (iii) party and play use-methamphetamine, gamma-hydroxybutyrate and amyl nitrites (i.e. poppers) with erectile dysfunction drugs (31%); and (iv) high polysubstance use (4%). SMM of colour and those with a history of incarceration were more commonly classified as engaging in methamphetamine and crack-cocaine use compared to party and play use. Men with higher sexual compulsivity scores were more commonly classified as engaging in party and play use and polysubstance use. There is substantial heterogeneity in polysubstance use patterns among SMM living with HIV who use methamphetamine. This will inform the development of tailored substance use interventions addressing the unique needs of SMM of colour and targeting sexual compulsivity as a prominent comorbidity for some men.

An Explicit Mean-Covariance Parameterization for Multivariate Response Linear Regression

We develop a new method to fit the multivariate response linear regression model that exploits a parametric link between the regression coefficient matrix and the error covariance matrix. Specifically, we assume that the correlations between entries in the multivariate error random vector are proportional to the cosines of the angles between their corresponding regression coefficient matrix columns, so as the angle between two regression coefficient matrix columns decreases, the correlation between the corresponding errors increases. We highlight two models under which this parameterization arises: a latent variable reduced-rank regression model and the errors-in-variables regression model. We propose a novel nonconvex weighted residual sum of squares criterion which exploits this parameterization and admits a new class of penalized estimators. The optimization is solved with an accelerated proximal gradient descent algorithm. Our method is used to study the association between microRNA expression and cancer drug activity measured on the NCI-60 cell lines. An R package implementing our method, MCMVR, is available online.

Relative contributions of health behaviours versus social factors on perceived and objective weight status in Canadian adolescents.

Public health interventions for adolescent "obesity prevention" have focused predominantly on individualistic health behaviours (e.g., diet and physical activity) at the expense of recognizing body weight diversity and the array of social factors (e.g., stigma and discrimination of marginalized identities) that may be linked to weight status. Research is needed to examine the extent to which individualistic health behaviours versus social factors contribute to weight status in adolescents. As such, the aim of this study was to investigate the relative contribution of individualistic health behaviours versus social factors to objective and perceptual indices of weight status. Cross-sectional survey data were collected as part of the Toronto Public Health Student Survey and comprised students 12 to 19years of age (N = 5515). Measures included perceived and objective weight status, social and demographic factors (e.g., gender, sexual orientation, school connectedness), and health behaviours (e.g., physical activity, nutritious consumption). Findings from latent variable regression models partially supported hypotheses, whereby social factors (i.e., age, sex, socio-economic access, sexual minority status) contribute similar amounts of variance, or relatively more variance in weight indices, compared to health behaviours (e.g., physical activity, nutritious consumption). Contrary to traditional views of adolescent weight status, physical activity (i.e., school-based, individual, active transport) and nutritious consumption (i.e., fruits, vegetables, milk) were not associated with weight status, when considering social factors. These findings challenge the utility of public health approaches that target individualistic behaviours as critical risk factors in "obesity prevention" efforts in adolescence.