Regression Estimates Research Articles

Deep learning-based estimation of time-dependent parameters in Markov models with application to nonlinear regression and SDEs

We present a novel deep-learning method for estimating time-dependent parameters in Markov processes through discrete sampling. Departing from conventional machine learning, our approach reframes parameter approximation as an optimization problem using the maximum likelihood approach. Experimental validation focuses on parameter estimation in multivariate regression and stochastic differential equations (SDEs). Theoretical results show that the real solution is close to SDE with parameters approximated using our neural network derived under specific conditions. Our work contributes to SDE-based model parameter estimation, offering a versatile tool for diverse fields.

Unlocking Green Productivity Through Green Lending, Digitalisation and Financial Literacy

Purpose — This study aims to measure green productivity in the agricultural sector, specifically on Sumatra Island, with some driving factors such as green lending from rural banks (both conventional and Sharīʿah), digitalisation, and the intermediation of financial literacy. Design/Methodology/Approach — The study employs panel data from 154 regencies on Sumatra Island for the period 2018 to 2022. Data Envelopment Analysis was employed to generate agricultural green total factor productivity, followed by regression estimation to see the influence of each determining factor. Findings — This study reveals the significant impact of green lending on green productivity. However, in Sharīʿah rural banks, which adhere strictly to Sharīʿah rules, this impact is positive only when moderated by financial literacy. On the other hand, digitalisation is also found to influence green productivity, but the effect weakened when the level of financial literacy could not be controlled. Originality/Value — To the best of the author’s knowledge, this study is the first to observe green productivity in the agricultural sector within a dual economic system that fosters both Islamic and conventional financial practices at the regional level. Practical Implications — This study implies that governments and banks should integrate green lending and digitalisation across various levels of financial literacy.

Determinants of Government Spending and Investment on District/ City Economic Growth in Jambi Province.

his research is entitled Determinants of Government Spending and Investment on the Economic Growth of Districts and Municipalities in Jambi Province. The study aims to determine and analyze the Determinants of Employee Expenditure, Capital Expenditure, Goods and Services Expenditure, PMDN, and PMA on District/City Economic Growth in Jambi Province in 2017-2022. The analysis method uses quantitative descriptive analysis with secondary data, the analysis tool uses panel data regression using cross-section data, namely the research objects of 11 districts and cities in the province, and time series data, namely 2017-2022 with a total number of observations of 66 objects, then using tests related to panel data regression, namely the selection of chow test models, Hausman tests, and Lagrange multiplier tests, and using statistical F and T hypothesis tests and coefficients of determination. The results showed that the regression estimation of panel data using the fixed effect model showed that the variables of capital expenditure, employee expenditure, and expenditure on goods and services had a significant effect on the economic growth of regencies and cities in Jambi Province. In contrast, the variables of foreign investment and domestic investment did not have a significant effect on the economic growth of Regencies and cities in Jambi Province.

Copula-based regression estimation in the presence of outliers

Ordinary least square estimators for linear regression models are highly influenced by outliers. Copula is a powerful tool for studying the dependence between response and predictor variables and may overcome several limitations associated with the classical linear regression models. In this study, we investigate the efficacy of the copula-based regression models over classical linear regression in the presence of outliers in the x, y, and x-y directions. We also examine the performance of several robust regression estimation techniques when variables are related via elliptical and Archimedean family of copulas and outliers are present in the data. Finally, real and artificial data sets are analyzed to compare the performance of various robust regression techniques.

Estimating substance use disparities across intersectional social positions using machine learning: An application of group-lasso interaction network.

An aim of quantitative intersectional research is to model the joint impact of multiple social positions on health risk behaviors. Although moderated multiple regression is frequently used to pursue intersectional research hypotheses, such parametric approaches may produce unreliable effect estimates due to data sparsity and high dimensionality. Machine learning provides viable alternatives, offering greater flexibility in evaluating many candidate interactions amid sparse data conditions, yet remains rarely employed. This study introduces group-lasso interaction network (glinternet), a novel machine learning approach involving hierarchical regularization, to assess intersectional differences in substance use prevalence. Utilizing variable selection and parameter stabilization functionality for main and interaction effects, glinternet was employed to examine two-way interactions between three primary social positions (gender, sexual orientation, and race) predicting heavy episodic drinking, cannabis use, and cigarette use prevalence. Analyses were conducted using the All of Us Research Program (N = 283,403), a national sample with high representation from populations historically underrepresented in biomedical research. Results were replicated using holdout cross-validation and compared against logistic regression estimates. Glinternet prevalence estimates were more stable across discovery and replication samples relative to logistic regression, particularly among sparsely represented groups. Prevalence estimates for cigarette and cannabis use were elevated among sexual minority and White cisgender women compared to heterosexual and non-White women, respectively. Glinternet may improve upon traditional moderated multiple regression methods for pursuing intersectional hypotheses by improving model parsimony and parameter stability, providing novel means for quantifying health disparities among intersectional social positions. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Entrepreneurial bricolage: A key to innovation for SMEs in a developing economy

Entrepreneurship promotes economic growth, particularly in developing economies where small and medium-sized enterprises (SMEs) are a significant source of employment and economic activity. However, SMEs in developing countries often face various resource constraints and weak institutions, forcing them to engage in entrepreneurial bricolage behavior by creatively combining existing resources. This study explores the relationship between entrepreneurial bricolage and product, process, and marketing innovation among SMEs in the Philippines. Logistic regression estimates reveal that entrepreneurial bricolage has a significant and positive impact on product, process, and marketing innovation. As entrepreneurial bricolage behavior increases, so does the probability of innovation, highlighting the importance of creative problem-solving in settings with inadequate resources. The study also emphasizes the need for policies that support SMEs by providing enabling resources, such as robust infrastructure and reliable communication platforms, to encourage firm innovation that fosters positive spill-over effects on the broader economy.

The moderating roles of healthcare expenditure in the fossil fuel consumption‐ mortality rate conflicts in fossil fuel‐dependent sub‐Saharan African countries

AbstractThis current study contributes to the extant literature by providing the first empirical evidence on the functional relationship between fossil fuel consumption, environmental pollution, and mortality rate in selected fossil fuels‐dependent nations in sub‐Saharan African countries with the moderating role of healthcare expenditures from 1982 to 2021. The empirical evidence relies on a battery of techniques comprising fully modified ordinary least squares, dynamic ordinary least squares, and panel quartile regression estimators. In order to establish sturdy empirical insights, fossil fuel is proxy by four indices including oil, coal, natural gas (at the disaggregated level), and fossil fuel (at the aggregated level). Mortality rate is also disaggregated into mortality rate, adult male, mortality rate, adult female, and infant mortality rate. Based on this disaggregation, the findings from the study reveal the following: First, the indicators for fossil fuel consumption have positive statistically significant impact on all three measures of mortality rate. Second, environmental pollution positively impacts the three indicators of mortality rate. Third, healthcare expenditure significantly reduces mortality rate, while its interaction with fossil fuel consumption moderates their unfavorable impacts on mortality rate. Fourth, with the exception of natural gas, the indicators of fossil fuel consumption and environmental pollution exert unfavorable impacts on mortality rate across all the quartiles. Emerging from these empirical findings, the study recommends promotion of cleaner sources of energy while at the same time improving healthcare expenditure as an interim measure pending full transition to renewable energy towards the attainment of a good health outcome in sub‐Saharan Africa countries.

Electrical Demand Analysis on Households and Industry in Indonesia

This research aims to determine the response of the household sector and industrial sector to electricity demand when there are changes in prices and income. The influence of price and income on electricity demand from both sectors can be seen through their elasticity. The approach used in this study is panel data from 33 provinces in Indonesia for the 2010-2020 time period. Panel data regression estimation techniques are used in this study to estimate the elasticity value. The results show that price elasticity in the household and industrial sectors is negative inelastic, but price does not significantly influence household electricity demand. Unlike price, income elasticity has a much higher value and positively and significantly influences electricity demand in both sectors. The number of customers, which reflects the increasing electrification ratio and population growth, significantly impacts electricity demand in the household and industrial sectors. Based on the results, it was found that the number of customers most influences electricity demand in the household sector. At the same time, income has the most significant influence on electricity demand in the industrial sector.

The Impact of Zoning Policy Implementation and the Education Production Function in Improving the Quality of Senior High School Education

This study aims to examine the inequality in the quality of education in 117 public high schools in DKI Jakarta, as well as the effect of implementing the PPDB zoning policy in improving the quality of school education. Utilizing a panel dataset of school units processed through quantitative analysis and Fixed-Effect (FE) econometric regression estimation, this study analyzes the Education Production Function. The research also investigates educational quality disparities based on Gini coefficients with the National Examination scores as indicators. The results of the study found that inequality between schools in DKI Jakarta was at a low level. Furthermore, it is found that the UN participant-to-teacher ratio, teacher education and experience, as well as the implementation of the zoning policy for student admission (PPDB), have an impact on improving school UN score achievements. In the future, school excellence will be distributed divergently.

The Moderating Effect of Vertical Agency Crisis on the Relationship Between Corporate Governance and Financial Performance in Nigeria

This study examines the moderating effect of vertical agency crises on the relationship between corporate governance mechanism and financial performance in Nigeria covering 2013 to 2022. To achieve this purpose, this study used precise proxies for the dependent, independent and moderator variables of interest. Based on a descriptive and ex-post facto research design, this study analyzed a set of panel data sourced from annual financial reports of eighteen (18) consumer goods firms drawn from a population of twenty-five (25) listed consumer goods firms in Nigeria. Mixed effect multi-level regression analysis estimator together with moderated regression analysis technique were employed to test the hypotheses after meeting necessary conditions for obtaining non-spurious least square regression estimates. The findings show that board ownership has a statistically significant relationship with financial performance, but board size, board independence and board gender diversity have insignificant relationship. Further, the findings from this study reveals among others that vertical agency crises have an insignificant moderating effect on the relationship between board independence and financial performance, the moderating effect is significantly positive in relation to board gender diversity for listed consumer goods firms in Nigeria . But for the variable of board size and board ownership, the moderating effect of vertical agency crises was seen to be statistically insignificant. Therefore, on the bases of the findings, this study suggests that managers should focus on mitigating vertical agency crises while concurrently strengthening board independence.

Socioeconomic Determinants of Insect Pest Control Methods’ Utilization for Storage in North Central Nigeria

The impetus for this study is the need to understand factors influencing the use of insect pest control methods in North Central Nigeria. Specifically, the study described the socioeconomic characteristics of respondents, assessed awareness and use of insect pest control methods, examined farmers’ considerations and investigated factors influencing insect pest control decisions. Using a structured interview schedule to elicit information, data was collected by means of multi-stage sampling technique to select a total of 384 cowpea farmers for this study. A purposive selection of Kwara and Niger States paved way for the purposive choice of two agricultural zones in Kwara state and one agricultural zone in Niger state. Percentages, frequency counts, ranks, mean, and ordinary least square regression was used to analyze the data. The study revealed that the intended time of sale, time required to apply method, and time of application were the factors most considered for store insect pest control. Ordinary Least Square regression estimates demonstrate positive significant relationship between farm size (t=4.681), group membership (t=4.688), and quantity stored in tons (t=2.312) and intensity of use. Conversely, years of education (t=-2.196), years of cowpea farming experience (t=-3.419), and frequency of extension contact (t=-.534) had significant negative relationship with intensity of use. In conclusion, it is recommended that policy makers and extension staff put into consideration farm size, group membership, quantity stored, years of education, years of cowpea farming experience and frequency of extension contact in optimizing extension strategy for promotion of indigenous and alternative insect pest control methods.

A naïve Bayes regularized logistic regression estimator for low-dimensional classification

To reduce the estimator's variance and prevent overfitting, regularization techniques have attracted great interest from the statistics and machine learning communities. Most existing regularized methods rely on the sparsity assumption that a model with fewer parameters predicts better than one with many parameters. This assumption works particularly well in high-dimensional problems. However, the sparsity assumption may not be necessary when the number of predictors is relatively small compared to the number of training instances. This paper argues that shrinking the coefficients towards a low-variance data-driven estimate could be a better regularization strategy for such situations. For low-dimensional classification problems, we propose a naïve Bayes regularized logistic regression (NBRLR) that shrinks the logistic regression coefficients toward the naïve Bayes estimate to provide a reduction in variance. Our approach is primarily motivated by the fact that naïve Bayes is functionally equivalent to logistic regression if naïve Bayes' conditional independence assumption holds. Under standard conditions, we prove the consistency of the NBRLR estimator. Extensive simulation and empirical experimental results show that NBRLR is a competitive alternative to various state-of-the-art classifiers, especially on low-dimensional datasets.

(104) PREDICTORS OF LENGTH OF STAY IN PELVIC INFLAMMATORY DISEASE HOSPITALIZATION: INSIGHTS FROM THE NATIONAL INPATIENT SAMPLE

Abstract Introduction Pelvic inflammatory disease (PID) is a common and potentially serious reproductive health condition that may require hospitalization for proper management. Accurate prediction of the length of stay (LOS) for PID hospitalizations can aid clinicians in resource allocation, patient counseling, and improving overall patient care. Prior studies have explored potential risk factors for a poor clinical course among women hospitalized with PID, but to our knowledge, no study to date has explored potential predictors of LOS in PID hospitalization. Objective This study aims to characterize the predictors for LOS in PID hospitalizations based on patient and hospital characteristics using a large national database. Methods The National Inpatient Sample (NIS) database was queried from 2019 to 2020 for all female patients 18 years and older with a primary diagnosis of acute pelvic inflammatory disease using ICD-10 codes. Data on patient demographics, clinical characteristics, comorbidities, and hospital-related variables were collected using variables as defined by NIS and summarized using descriptive statistics. Bivariate and multivariable associations between LOS and covariates were assessed using linear regression models and estimates were reported as regression coefficients. Covariates that were statistically significant in the bivariate models were included as covariates during the multivariate regression analysis. Results A total of 14,505 hospitalized female patients with a principal diagnosis code for acute PID were included. Mean LOS for PID hospitalization was 5.8 days, with a range of 0 – 88 days. The average patient age was 42.9 years (age range 18 – 90 years old), with 55.1% White, 23.4% Black, 12.5% Hispanic, and 2.3% Asian or Pacific Islander. The main primary payers included Medicaid (40.0%), private insurance (36.1%), Medicare (19.1%), and self-pay (4.7%). About 34% of patients were in the lowest quartile for median household income, 45% in the middle two quartiles, and 19% in the highest quartile. In univariable analysis, increased LOS was associated with increasing age, larger hospital bed size, urban teaching hospital status, and the presence of hypertension, diabetes, obesity, or immunodeficiency. Decreased LOS was associated with Hispanic or Native American race and having a non-Medicare primary payer. In multivariable analysis, increased LOS was associated with increasing age (+0.06 days per increasing year, p < 0.0001) and large hospital bed size (+0.6 days compared to small bed size, p = 0.03). Decreased LOS was associated with Asian or Pacific Islander race (-1.1 days compared to White race, p = 0.02), Native American race (-1.4 days compared to White race, p = 0.03), and being billed as ‘no charge’ (-2.1 days compared to Medicare, p = 0.03). Conclusions This study identifies possible predictors of LOS in PID hospitalization, providing insights that may help optimize resource allocation and better characterize factors that may impact PID management. Several factors, including age, race, expected primary payer, and bed size of hospital were found to significantly impact LOS. Additionally, this study highlights the potential role of social determinants of health in PID hospitalization and management and can help guide future research and initiatives aimed at improving population health and addressing disparities in PID. Disclosure No.

Poisson subsampling-based estimation for growing-dimensional expectile regression in massive data

Poisson subsampling-based estimation for growing-dimensional expectile regression in massive data

Exploring the policy implications of household wealth differentials and under-five child deaths in Ghana

BackgroundGhana was one of the few countries commended to have halved extreme poverty after the MDG era yet the under-five mortality rate (U5MR) was dire. Ghana’s U5MR is still above the national and global SDG target. As a result, the government has over the years implemented other complementary social and policy interventions to improve the socioeconomic status of households to ensure better child health outcomes. Yet key gaps exist and threaten child health outcome sustainability, and child health-related SDG targets. In this regard, we hypothesize that household wealth should therefore not influence under-five child deaths significantly.MethodsThe paper first reports the under-five mortality trend over the past 30 years in Ghana. It uniquely analyses the levels and trends in wealth as measured by the Gross National Income per capita at purchasing power parity (ln GNI/p, ppp) as a proxy of the mean Comparative Wealth Index (CWI) against under-five child deaths in Ghana. Using data from the 2014 Ghana Demographic and Health Survey and employing the logistic regression estimation technique, the paper further estimates the effect of household wealth and other key covariates on under-five child deaths. Implemented government poverty-alleviating policies and intervention programs to reduce childhood mortality in Ghana were reviewed to identify risk gaps to child deaths.ResultsThe 30 year trend analysis from 1990 to 2020 shows that the under-five mortality rate in Ghana is still high, falling short of the rate of decline that is expected toward achieving the 2030 SDG target for U5MR of 25 deaths per 1000 live births. Our empirical estimations show that household wealth still has a negative and significant relationship with under-five child deaths despite the government’s existing poverty-alleviating and pro-poor health policies. In addition, the results show that regional differences, maternal age, the number of children alive, and attendance to postnatal healthcare services are crucial for child survival.ConclusionThe results imply that the government should consolidate universal programmes with intensified targeted interventions toward improving household wealth, especially in the Northern Savannah regions. A new and dynamic child health policy is imminent to reflect strategies towards achieving the SDG target on U5MR. Additionally, intensive education on active maternal participation in postnatal care should be given critical consideration to improve child survival.

The role of firm complexity in the relationship between integrated reporting and earnings management

PurposeThis study aims to evaluate the relationship between integrated reporting and management’s opportunistic behavior (i.e., accrual and real earnings management) and the moderating role of firm complexity.Design/methodology/approachData of firms at the Johannesburg Stock Exchange were collected and analyzed. The Johannesburg Stock Exchange is currently the primary exchange that mandates the practice of integrated reporting. Regression estimation models and robustness tests were applied to the analysis.FindingsThis study concludes that integrated reporting quality reduces firms’ accrual and real earnings management practices. It further concludes that the significant negative effect of integrated reporting quality on firms’ earnings management practices is impeded by higher firm complexity.Originality/valueThis study enhances the literature on the behavioral effect of a combined financial and sustainability disclosure practice on both accrual and real earnings management, specifically targeting South Africa’s listed companies – the primary market currently mandates integrated reporting practice.

Uncertainty impacts on China’s agricultural commodity futures: a quantile perspective

ABSTRACT This study employs rolling quantile regression and non-parametric estimation methods, specifically Quantile-on-Quantile (QQ) and Causality-in-Quantiles (QC), to investigate the impact of uncertainty indices (Geopolitical Uncertainty (GPR), Climate Policy Uncertainty (CPU), and Economic Policy Uncertainty (EPU)) on China’s agricultural commodity futures. The rolling window quantile regression coefficients vary over time, with CPU having the most significant impact on agricultural commodity futures. The empirical results from the non-parametric estimation QQ method reveal when agricultural commodities are in a prosperous market, CPU and EPU exhibit relatively strong correlation, while GPR shows weaker or insignificant correlation. According to the non-parametric estimation QC method, the evidence suggests that the second-order (variance) causality relationship is relatively stronger than the first-order (mean) causality relationship, and the first-order mean causal relationship between CPU and corn and between EPU and wheat is not significant across the entire range of quantiles. The results on Hedging Effectiveness reveal that the efficiency of uncertainty hedging is substantiated. The hedging effectiveness of EPU exhibits improvement post-COVID-19, whereas the effectiveness of GPR and CPU is more pronounced in the pre-COVID-19 period.

Examining the relationship between environmental management accounting practices and return on equity in the South African chemical industry

Environmental management accounting practices (EMAPs) have become pervasive, and continued efforts to ensure universal implementation across various sectors often represent financial implications for organizations. Despite many studies that examined the relationship between EMAPs and financial performance, the debate is still inconclusive. Therefore, the study paves the way for chemical firms to explore the effectiveness of EMAPs’ implementation for both financial and environmental gain. The study used purposive sampling to gather quantitative secondary data from annual integrated reports of chemical firms to examine the relationship between EMAPs and financial performance in the South African chemical industries during 2016–2022. Following the results from the regression estimations, two of the EMAPs – water and energy usage – have had a positive relationship with financial performance, with the latter being highly significant. Contradictorily, carbon emissions and environmental expenditure adversely and insignificantly influenced financial performance. The results suggest that chemical firms have in place ineffective carbon management strategies that fail to generate sustainable returns. Overall, the results acknowledge the efforts of chemical industries in making substantial contributions to enhance environmental performance and encourage environmentalists and policymakers to reconfigure environmental policies for improved environmental and financial performance. Further research on environmental management accounting (EMA) barriers in chemical industries is imperative to achieving environmental sustainability.

The endogeneity of the natural rate of growth in Latin America

This study estimates the natural rate of growth à la León-Ledesma and Thirlwall for a panel of eleven Latin American countries from 2000 to 2023 using the seemingly unrelated regressions (SUR) estimator. Likewise, in the model, we incorporate a dummy variable in boom periods to determine whether endogeneity is present due to variations in the current rate of growth. Our results show that the behavior of the natural rate has a direct relationship with changes in the current rate; that is, it increases in boom periods and decreases during recessions, confirming the presence of endogeneity. In this sense, it is advisable to continue exploring the nature of the sensitivity of the natural growth rate to the economic situation. By providing recent evidence for Latin American countries with different degrees of economic development, our results allow us to conclude that the hypothesis of full employment on the supply side is not sufficient to explain the economic dynamics, so the inclusion of demand variables is necessary.

Allocation of the household food budget among shopping basket items: How is it influenced by promotions?

Retailers have been using promotion as a differentiation strategy that influences consumers' expenditures and their shopping basket budgetary allocation. This study assessed the effect of retail promotions on total shopping basket expenditure and determined whether promotions provoke a reallocation of the shopping budget. The analysis was performed on a chain of supermarkets in Catalonia, Spain using a consumer scanner data set from Kantar Worldpanel for 2017. The methodological approach had two steps: prediction of the effect of promotion on household expenditures using an expenditure regression model and estimation of the promotion own- and cross-effect using the censored Exact Affine Stone Index. Promotion had a positive own-effect and mostly a negative asymmetric cross-effect, implying a small but significant budget reallocation.