Bibliographic record
Abstract
In the absence of sufficiently large samples, due to expensive sampling costs or other constraints, statisticians can gain power for statistical inference by pooling the information of multiple available samples.If the underlying distributions of the samples being pooled are assumed to share some latent characteristics, one option at the statisticians' disposal is the semi-parametric Density Ratio Model (DRM).An area of interest concerning DRM inference is determining the pre-specified basis function of the model prior to estimating the model parameters -it is known that misspecification of this function can have adverse effects with respect to bias and mean-squared error of the estimates.The existing literature on current selection techniques for the basis function is limited.This thesis postulates a novel estimation method that applies the well known group lasso penalty to the Dual Empirical Likelihood (DEL) function.Analogous to its utility in regression, these penalized estimators are shown to generate sparse solutions that effectively simplify the model's basis function.Each penalty group is structured to contain coefficients of the same basis function term across all samples.An adaptive version of the estimator is also proposed in this thesis and shown to be both root-n consistent and selection consistent.As the group lasso penalty is non-differentiable at the origin, optimization of the objective function must be dealt with care.Three optimization algorithms are studied in application to the estimators: Subplex, Block Coordinate Gradient Descent, and Sequential Least Squares Quadratic Programming (SLSQP).SLSQP is observed to minimize the objective function sufficiently while converging in less iterations than the other candidates.The simulation study conducted shows that the proposed estimators are able to correctly select the basis function in certain scenarios.They display better performance when applied to data generated from symmetric distributions.The application of AIC and BIC to select the optimal tuning parameter value for the group lasso-based estimators are also shown to provide a modest level of utility in the model selection process.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.013 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".