Estimating Response-Maximized Decision Rules With Applications to Breastfeeding
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
To estimate the sequence of actions that optimizes response in a longitudinal setting, it is important to study the actions as part of a set of decision rules rather than in a series of single-action comparisons. We take as our motivating example the estimation of the set of decision rules for the duration of breastfeeding with a view to maximizing infant growth. Breastfeeding has many well-recognized beneficial effects on health and development. However observational evidence has suggested that breastfeeding is associated with reduced infant growth, although the long-term consequences for stature and adiposity remain controversial. The Promotion of Breastfeeding Intervention Trial (PROBIT) recruited 17,046 women in which hospitals and their affiliated polyclinics in Belarus were randomized to a breastfeeding promotion intervention or to standard care. In this article, we propose Structural Mean Models and estimate their parameters using G-estimation to obtain unbiased estimates of the effect of continued breastfeeding on infant growth (weight or length) at one year of age. We also implement a modified version of the G-estimation algorithm that is asymptotically unbiased; this is the first real-data application of the algorithm. Finally, we compare the decision rules implied by the G-estimates with the decision implied by a myopic optimization estimation approach, that is, we compare with decision rules that maximize response in the short-term. The breastfeeding regimes selected by each of the three models are optimal in the sense that specific criteria were optimized; the criteria considered here (maximizing weight or length) were chosen for simplicity, but may not lead to better overall health. We demonstrate in the context of a breastfeeding promotion intervention trial that optimal myopic decision strategies do not coincide with strategies that optimize a longer-term response. Please see the online supplements for a correction to this article.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it