expecon-referee-strategy
GitHub用于实验经济学论文投稿前，预判审稿人关于设计完整性、激励相容性及统计功效等核心异议的策略分析工具。旨在帮助作者提前识别弱点并制定防御或修改方案，确保稿件符合双盲评审标准。
Trigger Scenarios
Install
npx skills add brycewang-stanford/Awesome-Journal-Skills --skill expecon-referee-strategy -g -y
SKILL.md
Frontmatter
{
"name": "expecon-referee-strategy",
"description": "Use when anticipating the experimentalist referee's objections for an Experimental Economics (ExpEcon) manuscript before submission. Pre-empts the standard pushbacks; it does not draft the response letter (see expecon-rebuttal)."
}
Referee Strategy (expecon-referee-strategy)
When to trigger
- The paper is analytically done and you want to pre-empt the objections an ExpEcon referee will raise
- You suspect a specific weak point (a borderline procedure, modest power, a confound) and want to disarm it before submission
- You need to decide what to add now vs. defend in the letter later
- You want to read the draft the way two editors and two anonymous referees will
Who reviews you here, and what they look for
Experimental Economics uses anonymous (double-blind) refereeing, and accepted papers need the approval of two editors (检索于 2026-06；以官网为准). Your referees are practicing experimentalists. They are unimpressed by topical novelty and intensely focused on design integrity. They read in a predictable order: gates first, then the contrast, then inference, then reproducibility. Pre-empt each.
The standard pushbacks, ranked by lethality
- "This is deception / the incentives aren't real." Lethal and binary. If any procedure could be read as deceptive, name it explicitly and explain why it is not deception under the ESA definition (withholding ≠ misstating). Confirm every elicitation is incentive-compatible and report realized payments. Do not leave the referee to wonder.
- "It's underpowered / the unit of inference is wrong." The most common substantive rejection. Pre-empt with a power/MDE statement at the session/matching-group level and group-level inference. A precise null beats a noisy positive.
- "A confound, not the mechanism." Referees hunt for the alternative explanation: comprehension differences, session effects, order/learning, demand effects, experimenter effects. For each plausible alternative, show the design rules it out or report the test that does.
- "The treatment contrast is impure." If a treatment changes more than one thing, a referee will say the effect is unattributable. Show the minimal-pair logic, or add the control treatment that isolates the dimension.
- "The behavioral models aren't actually distinguished." If rival accounts predict the same data, the experiment is undiscriminating. Show where predictions diverge and that your design sits on that divergence.
- "Not reproducible / instructions missing." Have the instructions and z-Tree/oTree code ready at submission; referees check them for deception and comprehension.
- "Why here and not JEBO/GEB?" If a referee thinks the contribution is topical (JEBO) or theoretical (GEB), they may recommend transfer. Make the methods contribution explicit.
Turning anticipated objections into pre-emptive moves
- For each top objection, decide: fix now (add a treatment/analysis), disclose and bound (report the limit honestly with the test), or defend in text (a paragraph that pre-rebuts). Build a small internal table before submitting.
- Add a short "threats to validity" or "robustness" subsection that names the alternatives and the test for each — referees reward authors who do their job for them.
- Suggesting appropriate (non-conflicted) referees and clearly framing the methods contribution in the cover letter can shape who reviews and how.
A pre-submission self-review pass
Read your own draft in the referee's order and answer out loud:
- Gates — Could any sentence in Procedures be read as deception? Are all incentives real and incentive-compatible? Is this stated explicitly?
- Power — What is the MDE at the group level, and did the study have power to detect it? Would a referee call the n of groups (not subjects) thin?
- Confounds — List every alternative explanation a smart experimentalist would propose; for each, is there a design feature or test that kills it?
- Purity — Does each treatment manipulate exactly one thing?
- Discrimination — Do my contrasts actually separate the competing models, or do they all predict the same pattern?
- Repro — If a referee opened the appendix right now, are the instructions and code there and legible?
If any answer is weak, that is the item to fix or pre-empt before the desk editor ever sends it out.
The demand-effect and experimenter-effect family
Beyond the headline gates, experimentalist referees reliably probe a cluster of subtle validity threats — treat them as first-order, not nitpicks:
- Demand effects — did the design telegraph the hypothesis? Pre-empt with neutral framing, a post-experiment guess question, and (if needed) an obfuscated cover task.
- Experimenter effects — could the experimenter's presence or scripting bias responses? Use standardized on-screen instructions and double-blind payment where social-image is at stake.
- Order/learning effects — counterbalance and test, or justify a between-subject design.
- Selection into the lab — acknowledge the recruited pool and what it does and does not represent.
- House money / wealth effects — pay one random round; do not let accumulated earnings contaminate later choices.
For each, the winning move is the same: name the threat in a short validity subsection and show the design feature or test that addresses it, so the referee sees you raised it first.
Worked vignette (illustrative)
An auction experiment expects the "your stakes are hypothetical in the practice rounds" line to pass unnoticed. A self-review flags it as an incentive-compatibility gap a referee will pounce on. The pre-emptive move: state clearly that only practice rounds were unpaid, that all decision rounds were paid via one-randomly-selected-round, and report the realized average payment — turning a likely first-round objection into a non-issue before it is raised.
Shaping the cover letter and reviewer pool
Under double-blind review you cannot influence the referees' identities, but you can influence who the editors think should review it and how they frame the contribution. A cover letter that names the methods advance in one crisp sentence steers the paper toward experimentalist referees rather than topical ones. Suggesting two or three non-conflicted experts who know the procedure you use (not just the topic) helps the editors route the paper to readers who can evaluate the design on its merits — and who are less likely to mistake a methods paper for a JEBO-style topic paper.
Checklist
- The no-deception and incentive-compatibility defenses are explicit in the text, not implicit
- Power/MDE and group-level inference pre-empt the "underpowered/wrong unit" objection
- Each plausible confound (comprehension, session, order, demand, experimenter) has a design answer or a test
- The minimal-pair purity of each contrast is shown
- Rival behavioral models are shown to be distinguished by the design
- Instructions + experiment code ready for referees at submission
- The methods contribution is explicit enough to resist a "transfer to JEBO/GEB" suggestion
Anti-patterns
- Hoping a borderline procedure goes unnoticed instead of pre-empting the deception question
- A robustness section that lists checks but never names the alternative explanation each addresses
- Ignoring the unit-of-inference objection that nearly every experimentalist referee raises
- Leaving the methods contribution implicit, inviting a transfer recommendation
- Treating referee suggestions of demand effects as nitpicks rather than first-order threats
What the desk editor screens before referees ever see it
Two editors stand between you and acceptance, and a desk screen comes first. The fastest desk rejects at this journal are: a deception procedure; non-incentivized choices; missing instructions; and an obvious method mismatch (a topic paper that should be at JEBO, or a short note that should be at JESA). None of these are about quality — they are about fit and gates. Clear all four in the first read of your own draft, because no amount of clever analysis survives a procedure the editor reads as deceptive on page 6.
Output format
【Journal】Experimental Economics (ESA method flagship)
【Skill】expecon-referee-strategy
【Verdict】ready / pre-empt-first
【Top objections】ranked (gates → power/unit → confound → purity → models → repro)
【Per objection】fix-now / disclose-and-bound / defend-in-text
【Threats subsection】present and names each alternative + test? [Y/N]
【Methods contribution】explicit enough to resist transfer? [Y/N]
【Next skill】expecon-submission
Version History
- 1839142 Current 2026-07-05 13:13
