Effect Estimation Research Articles

Bayesian doubly robust estimation of causal effects for clustered observational data

Observational data often exhibit clustered structure, which leads to inaccurate estimates of exposure effect if such structure is ignored. To overcome the challenges of modelling the complex confounder effects in clustered data, we propose a Bayesian doubly robust estimator of causal effects with random intercept BART to enhance the robustness against model misspecification. The proposed approach incorporates the uncertainty in the estimation of the propensity score, potential outcomes and the distribution of individual-level and cluster-level confounders into the exposure effect estimation, thereby improving the coverage probability of interval estimation. We evaluate the proposed method in the simulation study compared with frequentist doubly robust estimators with parametric and nonparametric multilevel modelling strategies. The proposed method is applied to estimate the effect of limited food access on the mortality of cardiovascular disease in the senior population.

The Impact of Export Diversification on Economic Growth by Provinces in Türkiye

Beyond being a national measure of success, export diversification is closely linked to regional development, local employment, innovation capacity, and economic resilience. In the current study, we investigate the impact of export diversification on economic growth by provinces in the case of Türkiye for the period 2004-2018. For the purpose of the study, we employ static panel data models using Fixed Effects (FE) and Random Effects (RE) estimators, which account for unobserved heterogeneity. The findings of the study show that there is a significant negative relationship between per capita income and export intensity (HHI and GINI). In other words, as export diversification increases, per capita incomes increase. On the other hand, the coefficients obtained for the THEIL index are statistically insignificant. Therefore, it can be argued that indices measuring different dimensions of export diversification reflect different effects despite their conceptual similarities. The findings suggest that promoting export diversification can enhance regional economic growth and resilience, indicating that policymakers should prioritize strategies that encourage diversification of export portfolios, especially in underdeveloped regions, to foster balanced economic growth.

Động cơ tích lũy dự trữ ngoại hối: Bằng chứng thực nghiệm tại các quốc gia đang phát triển

Using two way fixed effects estimation with data from 77 developing countries for the period 2000 2022, the article provides evidence that foreign exchange reserve accumulation comes from both motives: mercantilism and insurance, in which the insurance motive has the greater influence. The exchange rate regime and the level of financial openness are two factors that change the level of influence of accumulative motives of foreign reserves. The impact of exchange rates on foreign exchange can be in opposite directions among different exchange rate regimes. Countries with high financial openness are often more concerned with the internal liquidity reserve motive, on the contrary, countries with low financial openness are more likely to accumulate reserves for mercantilism. The research results provide an important policy implication for developing countries, including Vietnam, on the need to calculate and maintain the optimal scale of foreign reserve to ensure multiple management goal, and at the same time consider the characteristics of the exchange rate regime and the openness of the capital account of the economy itself.

Long-term employment effects of the minimum wage in Germany: New data and estimators

We investigate the long-term effects of the introduction of the German minimum wage in 2015 and its subsequent increases on regional employment. Using comprehensive survey data, we are able to measure the regional bite of the minimum wage in 2014, just before its introduction, as well as in 2018, before it was raised substantially in several steps. The introduction mainly affected the labour market in East Germany, while the minimum wage increases increasingly affected low-wage regions in West Germany, with about one third of regions changing their (binary) treatment status between 2014 and 2018. We use different specifications and extensions of the canonical difference-in-differences approach, as well as a set of new estimators that allow unbiased effect estimation with a staggered treatment adoption and heterogeneous treatment effects. Our results show a small negative effect on total dependent employment of 0.5%, driven by a significant reduction in marginal employment of 2.4%. The extended specifications suggest additional effects of the minimum wage increases, as well as stronger negative effects for those regions that were strongly affected by the minimum wage in both periods.

Using Quantitative Bias Analysis to Adjust for Misclassification of COVID-19 Outcomes: An Applied Example of Inhaled Corticosteroids and COVID-19 Outcomes.

During the pandemic, there was concern that underascertainment of COVID-19 outcomes may impact treatment effect estimation in pharmacoepidemiologic studies. We assessed the impact of outcome misclassification on the association between inhaled corticosteroids (ICS) and COVID-19 hospitalisation and death in the United Kingdom during the first pandemic wave using probabilistic bias analysis (PBA). Using data from the Clinical Practice Research Datalink Aurum, we defined a cohort with chronic obstructive pulmonary disease (COPD) on 1 March 2020. We compared the risk of COVID-19 hospitalisation and death among users of ICS/long-acting β-agonist (LABA) and users of LABA/LAMA using inverse probability of treatment weighted (IPTW) logistic regression. We used PBA to assess the impact of non-differential outcome misclassification. We assigned beta distributions to sensitivity and specificity and sampled from these 100 000 times for summary-level and 10 000 times for record-level PBA. Using these values, we simulated outcomes and applied IPTW logistic regression to adjust for confounding and misclassification. Sensitivity analyses excluded ICS + LABA + LAMA (triple therapy) users. Among 161 411 patients with COPD, ICS users had increased odds of COVID-19 hospitalisations and death compared with LABA/LAMA users (OR for COVID-19 hospitalisation 1.59 (95% CI 1.31-1.92); OR for COVID-19 death 1.63 (95% CI 1.26-2.11)). After IPTW and exclusion of people using triple therapy, ORs moved towards the null. All implementations of QBA, both record- and summary-level PBA, modestly shifted the ORs away from the null and increased uncertainty. We observed increased risks of COVID-19 hospitalisation and death among ICS users compared to LABA/LAMA users. Outcome misclassification was unlikely to change the conclusions of the study, but confounding by indication remains a concern.

All-breed single-step genomic best linear unbiased predictor evaluations for fertility traits in US dairy cattle.

The US dairy cattle genetic evaluation is currently a multistep process, including multibreed traditional BLUP estimations followed by single-breed SNP effects estimation. Single-step GBLUP (ssGBLUP) combines pedigree and genomic data for all breeds in one analysis. Unknown parent groups (UPG) or metafounders (MF) can be used to address missing pedigree information. Fertility traits are notably difficult to evaluate due to low heritabilities, changing management, and a higher recent emphasis on selection to move in a favorable direction. We assessed bias, dispersion, and accuracy of fertility traits in all-breed US dairy cattle using pedigree-based BLUP (PBLUP) and ssGBLUP with UPG or MF; with 5% or 10% residual polygenic effect. Validation methods included the linear regression method and comparison of early and late deregressed proofs for Holstein and Jersey breeds. By comparing MF or UPG in PBLUP, we observed similar results in terms of bias, dispersion, and correlations between early and recent predictions. When genomics was used, ssGBLUP with MF and 10% residual polygenic effect consistently outperformed other models regarding bias, dispersion, and correlations. Compared with multistep results, ssGBLUP with MF and 10% residual polygenic effect showed less bias and increased correlations but slightly overdispersed estimates. Overall, genomic prediction of fertility traits using ssGBLUP was accurate and unbiased, more so with MF than with UPG.

Double Sampling for Informatively Missing Data in Electronic Health Record-Based Comparative Effectiveness Research.

Missing data arise in most applied settings and are ubiquitous in electronic health records (EHR). When data are missing not at random (MNAR) with respect to measured covariates, sensitivity analyses are often considered. These solutions, however, are often unsatisfying in that they are not guaranteed to yield actionable conclusions. Motivated by an EHR-based study of long-term outcomes following bariatric surgery, we consider the use of double sampling as a means to mitigate MNAR outcome data when the statistical goals are estimation and inference regarding causal effects. We describe assumptions that are sufficient for the identification of the joint distribution of confounders, treatment, and outcome under this design. Additionally, we derive efficient and robust estimators of the average causal treatment effect under a nonparametric model and under a model assuming outcomes were, in fact, initially missing at random (MAR). We compare these in simulations to an approach that adaptively estimates based on evidence of violation of the MAR assumption. Finally, we also show that the proposed double sampling design can be extended to handle arbitrary coarsening mechanisms, and derive nonparametric efficient estimators of any smooth full data functional.

Epidemiology of Rounding Error.

This work represents a significant contribution to understanding the importance of appropriately rounding numbers with minimal error. That is, to reduce inexact rounding and data truncation error and simultaneously eliminate unintentional misleading findings in epidemiological studies. The rounding of numbers represents a compromise solution that attempts to find a balance between the loss of information from reporting too few significant digits versus retaining more digits than necessary. Substituting a rounded number for its original value may be acceptable and practical in many applied situations if an adequate degree of accuracy is retained. On the other hand, numeric error may result from improper rounding or data truncation which, in effect, compromises the credibility of study findings and may lead to a false sense of discovery. Performing complex computations on such values, especially when sequential or composite operations are involved, can lead to error propagation and inaccurate results. Having an overall awareness of the nature and impact of rounding error, including preventive actions, can contribute greatly to the integrity of research, yielding more reliable and accurate conclusions. Heuristic examples are provided to illustrate the consequences of rounding and data truncation error in epidemiology studies, specifically those pertaining to relative effect estimation.

Improving randomized controlled trial analysis via data-adaptive borrowing

SUMMARY In recent years, real-world external controls have grown in popularity as a tool to empower randomized placebo-controlled trials, particularly in rare diseases or cases where balanced randomization is unethical or impractical. However, as external controls are not always comparable to the trials, direct borrowing without scrutiny may heavily bias the treatment effect estimator. Our paper proposes a data-adaptive integrative framework capable of preventing unknown biases of the external controls. The adaptive nature is achieved by dynamically sorting out a comparable subset of the external controls via bias penalization. Our proposed method can simultaneously achieve (a) the semiparametric efficiency bound when the external controls are comparable and (b) selective borrowing that mitigates the impact of the existence of incomparable external controls. Furthermore, we establish statistical guarantees, including consistency, asymptotic distribution, and inference, providing type-I error control and good power. Extensive simulations and two real-data applications show that the proposed method leads to improved performance over the trial-only estimator across various bias-generating scenarios.

Impact of Innovation and Trade Participation on Economic Growth Among Selected African Countries

This study examines the impact of innovation and trade participation on economic growth among selected African countries from 1996 to 2021. This study applied nonstationary heterogeneous panel models utilizing pooled mean group estimators, mean group estimators and dynamic fixed effects estimators. Based on the results of the Hausman test, the PMG estimator was considered for the data analysis. The study revealed the significant role of industrial design, patents, trademarks, research and development, exports, and balanced trade policies in driving long-term economic growth in Africa. The synergies between innovation and trade participation are studied, emphasizing their amplified impact on economic growth. The study suggests practical policy recommendations for the selected African countries, urging governments to establish innovation ecosystems, encourage investments in design, strengthen intellectual property protection, prioritize infrastructure development, and develop balanced trade policies. Focusing on SMEs, offering incentives for research and development, and establishing trade promotion agencies are also recommended to create a conducive environment for innovation, trade, and sustained economic growth.

Bank Risk Taking Behaviour and Performance: Evidence from Indonesian Bank

This study aim to assess the impact of risk management and internal control on Indonesian Banks' risk-taking behaviour and performance. By assessing sample of 60 Banks spanning from 2013 and 2022 using the Generalized Least Square random effect estimator model, results show that while risk management and internal control lower credit risk and operational risk, they have no significant impact on liquidity risk. This impact is more pronounced for Government owned Banks than for privately owned Banks. Furthermore, the research demonstrates that risk management and internal control improve Banks' performance particularly in ROA and ROE parameters.

Quantile outcome adaptive lasso: Covariate selection for inverse probability weighting estimator of quantile treatment effects.

When using the propensity score method to estimate the treatment effects, it is important to select the covariates to be included in the propensity score model. The inclusion of covariates unrelated to the outcome in the propensity score model led to bias and large variance in the estimator of treatment effects. Many data-driven covariate selection methods have been proposed for selecting covariates related to outcomes. However, most of them assume an average treatment effect estimation and may not be designed to estimate quantile treatment effects (QTEs), which are the effects of treatment on the quantiles of outcome distribution. In QTE estimation, we consider two relation types with the outcome as the expected value and quantile point. To achieve this, we propose a data-driven covariate selection method for propensity score models that allows for the selection of covariates related to the expected value and quantile of the outcome for QTE estimation. Assuming the quantile regression model as an outcome regression model, covariate selection was performed using a regularization method with the partial regression coefficients of the quantile regression model as weights. The proposed method was applied to artificial data and a dataset of mothers and children born in King County, Washington, to compare the performance of existing methods and QTE estimators. As a result, the proposed method performs well in the presence of covariates related to both the expected value and quantile of the outcome.

Digital infrastructure and cognitive ability of children

Digital technologies have become an integral part of most children’s lives, significantly influencing their development. This study takes the Broadband China Pilot Policy as a quasi-experiment, using the data from China Family Panel Studies (CFPS) to analyze the effects of digital infrastructure on children’s cognitive ability. The extended two-way fixed effects estimator was employed to conduct the Staggered Difference-in-Differences estimation. The findings indicate that the average treatment effect of treatment group of “Broadband China” is significantly positive on children’s cognitive ability. Additionally, heterogeneities of gender and urban-rural were found: (1) The Broadband China policy had a significant positive impact on boys only; (2) The policy had a greater and more significant impact on the word test scores results of urban children; (3) The policy had a negative impact on the math test scores of urban children, while showing a positive impact on the math test scores of rural children. Finally, the paper makes the following recommendations: (1) Digital infrastructure development should be emphasized; (2) More emphasis should be placed on rural areas when building digital infrastructure; (3) Gender differences should be considered when formulating policies to help girls benefit from digital technologies.

Identifying causal genotype-phenotype relationships for population-sampled parent-child trios.

The process by which genes are transmitted from parent to child provides a source of randomization preceding all other factors that may causally influence any particular child phenotype. Because of this, it is natural to consider genetic transmission as a source of experimental randomization. In this work, we show how parent-child trio data can be leveraged to identify causal genetic loci by modeling the randomization during genetic transmission. We develop a new test, the transmission mean test (TMT), together with its unbiased estimator of the average causal effect, and derive its causal properties within the potential outcomes framework. We also prove that the transmission disequilibrium test (TDT) is a test of causality as a complementary case of the TMT for the affected-only design. The TMT and the TDT differ in the types of traits that they can handle and the study designs for which they are appropriate. The TMT handles arbitrarily distributed traits and is appropriate when trios are randomly sampled; the TDT handles dichotomous traits and is appropriate when sampling is based on a child's trait status. We compare the transmission-based methods with established approaches for genotype-phenotype analyses to clarify conditions appropriate for each method, what conclusions can be drawn by each one, and how these methods can be used together.

Multiplicative versus additive modelling of causal effects using instrumental variables for survival outcomes - a comparison.

Instrumental variables (IVs) methods have recently gained popularity since, under certain assumptions, they may yield consistent causal effect estimators in the presence of unmeasured confounding. Existing simulation studies that evaluate the performance of IV approaches for time-to-event outcomes tend to consider either an additive or a multiplicative data-generating mechanism (DGM) and have been limited to an exponential constant baseline hazard model. In particular, the relative merits of additive versus multiplicative IV models have not been fully explored. All IV methods produce less biased estimators than naïve estimators that ignore unmeasured confounding, unless the IV is very weak and there is very little unmeasured confounding. However, the mean squared error of IV estimators may be higher than that of the naïve, biased but more stable estimators, especially when the IV is weak, the sample size is small to moderate, and the unmeasured confounding is strong. In addition, the sensitivity of IV methods to departures from their assumed DGMs differ substantially. Additive IV methods yield clearly biased effect estimators under a multiplicative DGM whereas multiplicative approaches appear less sensitive. All can be extremely variable. We would recommend that survival probabilities should always be reported alongside the relevant hazard contrasts as these can be more reliable and circumvent some of the known issues with causal interpretation of hazard contrasts. In summary, both additive IV and Cox IV methods can perform well in some circumstances but an awareness of their limitations is required in analyses of real data where the true underlying DGM is unknown.

Common correlated effects estimation of nonlinear panel data models

Abstract This paper focuses on estimating the coefficients and average partial effects of observed regressors in nonlinear panel data models with interactive fixed effects, using the common correlated effects (CCE, hereafter) framework. The proposed two-step estimation method involves applying principal component analysis to estimate the latent factors based on cross-sectional averages of the regressors in the first step, and jointly estimating the coefficients of the regressors and the factor loadings in the second step. The asymptotic distributions of the proposed estimators are derived under general conditions, assuming that the number of time-series observations is comparable to the number of cross-sectional observations. To correct for asymptotic biases of the estimators, we introduce both analytical and split-panel jackknife methods, and confirm their good performance in finite samples using Monte Carlo simulations. Finally, the proposed method is used to study the arbitrage behavior of nonfinancial firms across different security markets.

Towards Robust Causal Inference in Epidemiological Research: Employing Double Cross-fit TMLE in Right Heart Catheterization Data.

Within epidemiological research, estimating treatment effects from observational data presents notable challenges. Targeted Maximum Likelihood Estimation (TMLE) emerges as a robust method, addressing these challenges by accurately modeling treatment effects. This approach uniquely combines the precision of correctly specified models with the versatility of data-adaptive, flexible machine learning algorithms. Despite its effectiveness, TMLE's integration of complex algorithms can introduce bias and under-coverage. This issue is addressed through the Double Cross-fit TMLE (DC-TMLE) approach, enhancing accuracy and reducing biases inherent in observational studies. However, DC-TMLE's potential remains underexplored in epidemiological research, primarily due to the lack of comprehensive methodological guidance and the complexity of its computational implementation. Recognizing this gap, our paper contributes a detailed, reproducible guide for implementing DC-TMLE in R, aimed specifically at epidemiological applications. We demonstrate the utility of this method using an openly available clinical dataset, underscoring its relevance and adaptability for robust epidemiological analysis. This guide aims to facilitate broader adoption of DC-TMLE in epidemiological studies, promoting more accurate and reliable treatment effect estimations in observational research.

Impact of Women's Empowerment on Female Account Ownership: A Case Study of Selected Sub-Saharan African Countries (2011 – 2021)

This study investigated women's empowerment and its impact on female account ownership, using selected sub-Saharan African countries as a case study from 2011 to 2021. The central research question covers whether women's empowerment has any impact on the financial inclusion of women in sub-Saharan Africa, with an emphasis on account ownership by women as a yardstick to measure financial inclusion. The selected women empowerment variables comprised the percentage of the total labour force, namely female labour force (FLF), equality in the rights to ownership of property (EOR), non-discrimination of women in employment (NDW), choice of where to live by women (CWL), mean years of schooling of females (MYS), and women political empowerment (WPE). By means of pooled ordinary least square regression, using the fixed effects estimation model applied to the panel data from 38 sub-Saharan African (SSA) countries with available data for the period 2011 to 2021, as reported on the three-yearly database, this research was able to establish the impact of some of the selected variables of women empowerment on female account ownership in SSA. This study hinges on the undeniable potential of women to lift Africa from multi-dimensional poverty since a greater proportion of household consumption is contributed by the income of women rather than men in Africa. The authors, therefore, recommend, among others, that African governments should invest in policies that will encourage gender equality and women's emancipation to boost women's empowerment, which is the gateway to enhanced financial inclusion of African women and the attendant gain of accelerated poverty reduction in Africa.

TEMR: Trans-ethnic mendelian randomization method using large-scale GWAS summary datasets.

Available large-scale genome-wide association study (GWAS) summary datasets predominantly stem from European populations, whilesample sizes for other ethnicities, notably Central/South Asian, East Asian, African, Hispanic, etc., remain comparatively limited, resulting in low precision of causal effect estimations within these ethnicities when using Mendelian randomization (MR). In this paper, we propose a trans-ethnic MR method, TEMR, to improve the statistical power and estimation precision of MR in a target population that is underrepresented, using trans-ethnic large-scale GWAS summary datasets. TEMR incorporates trans-ethnic genetic correlation coefficients through a conditional likelihood-based inference framework, producing calibrated p values with substantially improved MR power. In the simulation study, compared with other existing MR methods, TEMR exhibited superior precision and statistical power in causal effect estimation within the target populations. Finally, we applied TEMR to infer causal relationships between concentrations of 16 blood biomarkers and the risk of developing five diseases (hypertension, ischemic stroke, type 2 diabetes, schizophrenia, and major depression disorder) in East Asian, African, and Hispanic/Latino populations, leveraging biobank-scale GWAS summary data obtained from individuals of European descent. We found that the causal biomarkers were mostly validated by previous MR methods, and we also discovered 17 causal relationships that were not identified using previously published MR methods.

Efficient designs for three-sequence stepped wedge trials with continuous recruitment.

The standard approach to designing stepped wedge trials that recruit participants in a continuous stream is to divide time into periods of equal length. But the choice of design in such cases is infinitely more flexible: each cluster could cross from the control to the intervention at any point on the continuous time-scale. We consider the case of a stepped wedge design with clusters randomised to just three sequences (designs with small numbers of sequences may be preferred for their simplicity and practicality) and investigate the choice of design that minimises the variance of the treatment effect estimator under different assumptions about the intra-cluster correlation. We make some simplifying assumptions in order to calculate the variance: in particular that we recruit the same number of participants, , from each cluster over the course of the trial, and that participants present at regularly spaced intervals. We consider an intra-cluster correlation that decays exponentially with separation in time between the presentation of two individuals from the same cluster, from a value of for two individuals who present at the same time, to a value of for individuals presenting at the start and end of the trial recruitment interval. We restrict attention to three-sequence designs with centrosymmetry - the property that if we reverse time and swap the intervention and control conditions then the design looks the same. We obtain an expression for the variance of the treatment effect estimator adjusted for effects of time, using methods for generalised least squares estimation, and we evaluate this expression numerically for different designs, and for different parameter values. There is a two-dimensional space of possible three-sequence, centrosymmetric stepped wedge designs with continuous recruitment. The variance of the treatment effect estimator for given and can be plotted as a contour map over this space. The shape of this variance surface depends on and on the parameter , but typically indicates a broad, flat region of close-to-optimal designs. The 'standard' design with equally spaced periods and 1:1:1 allocation rarely performs well, however. In many different settings, a relatively simple design can be found (e.g. one based on simple fractions) that offers close-to-optimal efficiency in that setting. There may also be designs that are robustly efficient over a wide range of settings. Contour maps of the kind we illustrate can help guide this choice. If efficiency is offered as one of the justifications for using a stepped wedge design, then it is worth designing with optimal efficiency in mind.