Statistical Inference in Semiparametric Regression Models

Semiparametric regression models combine the interpretability of parametric components with the flexibility of nonparametric functions, enabling analysts to capture complex relationships without imposing overly restrictive assumptions. Inference in these models often centres on estimating finite-dimensional parameters—such as regression coefficients—and constructing valid confidence intervals or hypothesis tests for them, while simultaneously modelling unknown smooth functions or nuisance elements. Key frameworks include partially linear models, which embed a nonparametric term alongside a linear predictor; varying-coefficient models, where coefficients themselves are smooth functions of covariates; and single-index models, which reduce multivariate covariates to a univariate index before applying a link function. Advances in this field address challenges arising from high dimensionality, endogeneity, heteroscedasticity and missing data, often through novel optimisation schemes or likelihood-based techniques that preserve asymptotic efficiency. Practical applications span economics, epidemiology and environmental science, where semiparametric methods balance parsimony and adaptivity to deliver robust insight into real-world phenomena.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Statistical Inference in Semiparametric Regression Models publication trend

The graph below shows the total number of articles in statistical inference in semiparametric regression models across all publications each year (not limited to Nature Index journals).

Technical terms

Semiparametric regression model: A statistical model combining parametric and nonparametric elements, balancing structure and flexibility.

Partially linear model: A semiparametric form with a linear term for some covariates and an unspecified smooth function for others.

Varying-coefficient model: A regression framework in which coefficients depend smoothly on one or more covariates.

Empirical likelihood: A nonparametric likelihood method that constructs confidence regions without specifying a full distribution.

Alternating-direction multiplier method (ADMM): An optimisation algorithm that decomposes complex problems into simpler subproblems.

Instrumental variable: An external variable used to address endogeneity by isolating exogenous variation in a covariate.

Heteroscedasticity: A condition in which the variability of errors differs across observations, affecting standard inference.

References

1. Application of LADMM and As-LADMM for a High-Dimensional Partially Linear Model. Mathematics (2023).
2. Estimation in Semi-Varying Coefficient Heteroscedastic Instrumental Variable Models with Missing Responses. Mathematics (2023).
3. Orthogonality based modal empirical likelihood inferences for partially nonlinear models. AIMS Mathematics (2024).