FlowingFlowing
← Back to all development cases
Development case · blind model judgment pending independent human review; not a formal Benchmark conclusion.
AutocompleteEvidence version preferred37 / 162 · 9c6ee2b1d7a062c8

Causal inference methods for combining randomized trials and observational studies: a review

Statistics · 2011.08047v4

FLOWING EVIDENCE BENCHMARK

How do we tell whether Evidence helps?

At the same writing position, with the same model and task, how does supplying retrieved paper passages change the first output? We compare matched versions and retain ties, unusable outputs and incomplete reviews.

SAME DRAFT · EVIDENCE ON OR OFF

01 · FIXED WRITING POSITION

MANUSCRIPT

Same writing position ▌

Matched autocomplete at the same draft position

02 · TWO MATCHED INPUTS

SHARED BY BOTH

Manuscript context, model, task and prompt

A · With Evidence

Retrieved paper passages supplied

B · Without Evidence

No retrieved passages supplied

LLM

Same model and version

A → first output

B → first output

Blind judge agent

First outputs are anonymized as X and Y

Continuation review: accuracy, fit to the writing task and usability
Returns: X preferred / tie / Y preferred / both unusable

Order check: X / Y → Y / X

HOW DOES BLIND JUDGING WORK?

① Anonymize both outputs
The judge sees the same draft and both first outputs without knowing which received Evidence.

② Compare and swap order
The judge applies task-specific criteria in X/Y and then Y/X order.

③ Review disagreements
A third pass resolves disagreements. Incomplete reviews remain in the denominator.

How are autocomplete positions stratified?

Before seeing generation outcomes, we check whether a retrieved passage contains a specific proposition that directly supports the next writing move. Those positions appear in the left opportunity group; the rest are ordinary positions on the right. We select a balanced sample from admitted papers in each field. The 50/50 split is experimental, not a measure of how often either type occurs in writing.

AUTOCOMPLETE · 110

Biology, statistics and astrophysics

55 positions on each side; statistics uses the ten-paper rerun.

AUTOCOMPLETE · 52

Psychology and climate science

26 positions on each side; psychology includes nine papers and climate science four.

How are the table percentages calculated?

Across five fields there are 81 source-opportunity positions. The original blind review preferred Evidence in 51; a task check moved one empty Evidence continuation to both unusable, leaving 50 in public counts. In the ordinary group, another pair of empty outputs moved from no winner to both unusable. Original verdicts remain visible on case pages.

50Evidence version preferred
÷
81All positions in this group
=
62%Evidence preference in this group

Source contribution is a separate review: 16 of 21 Evidence wins entered into source review directly used retrieved papers; another 29 wins await review.

These are development-stage model judgments pending independent human review. They are not formal Benchmark conclusions and do not, on their own, establish causality.

Manuscript writing position

Text excerpt · not a PDF page

Research manuscript · excerpt

Causal inference methods for combining randomized trials and observational studies: a review

1 Introduction

Keywords: Causal effect generalization; transportability; double robustness; data fusion; heterogeneous data; S-admissibility.

…n criteria which can lead to a trial sample that differs markedly from the population potentially eligible for the treatment. Therefore, the findings from RCTs can lack generalizability to a target population of interest. This concern is related to the aim of external validity , central in medical research ( Concato et al.,, 2000 ; Rothwell,, 2005 ; Green and Glasgow,, 2006 ; Frieden,, 2017 ) policy research ( Martel Garcia and Wantchekon,, 2010 ; Deaton and Cartwright,, 2018 ; Deaton et al.,, 2019 ; Jeong and Namkoong,, 2022 ) , and other fields such as advertising ( Gordon et al.,, 2019 ) .

↓ Next paragraph was generated here

The highlighted text is near the writing position. This is a layout of the evaluation input, not a PDF screenshot.

A matched comparison at the same writing positionJudgment: Evidence version preferred

Both arms use the same model, task and manuscript context. Only the retrieved passages differ. The first recorded output from each arm is shown below.

Evidence

Same model and position · retrieved passages supplied · first output

Preferred by blind review

When trial participation depends on observed covariates, propensity-score-based approaches can reweight the trial sample to resemble a target population, thereby generalizing estimated treatment effects beyond the original study sample.

Passages supplied to the Evidence version

On variance estimation for generalizing from a trial to a target population ↗

Randomized controlled trials (RCTs) provide strong internal validity compared with observational studies. However, selection bias threatens the external validity of randomized trials. Thus, RCT results may not apply to either broad public policy populations or narrow populations, such as specific in…

Read full passage excerpt

Randomized controlled trials (RCTs) provide strong internal validity compared with observational studies. However, selection bias threatens the external validity of randomized trials. Thus, RCT results may not apply to either broad public policy populations or narrow populations, such as specific insurance pools. Some researchers use propensity scores (PSs) to generalize results from an RCT to a target population. In this scenario, a PS is defined as the probability of participating in the trial conditioning on observed covariates. We study a model-free inverse probability weighted estimator (IPWE) of the average treatment effect in a target population with data from a randomized trial.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

A parallel concern arises from the internal validity of observational studies, which are often used to complement or substitute for trial evidence but are susceptible to confounding by indication and unmeasured variables. Combining the two designs therefore requires methods that can exploit the strengths of each while addressing their respective biases.

Judgment and task check

The blinded model review preferred the Evidence version. Its original rationale appears below.

Read the original model judgment

In this rationale, Output A = Evidence; Output B = No Evidence.

Output A is directly grounded in the source: it discusses propensity-score-based approaches for reweighting trial samples to target populations when trial participation depends on observed covariates, matching the source's description of PS-defined participation probability and IPWE for generalizing to target populations. It coherently continues the draft's theme of RCT generalizability concerns. Output B introduces a 'parallel concern' about observational studies' internal validity and confounding by indication/unmeasured variables, then claims combining designs 'requires methods that can exploit the strengths of each while addressing their respective biases.' This shifts to data fusion/combination, which is not discussed in the source. The source only discusses generalizing from RCTs to target populations using propensity scores, not combining RCTs with observational studies or their respective biases. The draft mentions 'data fusion' in keywords but the source provided does not support claims about combining designs or observational study weaknesses.

The passages are input material; their presence does not mean an output used them or that they support every claim. Draft and source passages are excerpted; no simulated scores or PDF appear here.