[ { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "The plan says the per-association results go in results/ but not how they are written. code/run now ends with code/report.R, added after registration, which writes results/associations.csv, R2.json, and R3.json by rounding what code/main.R computed and adding each paper's labels from data/frame.csv. It changes no analysis: code/main.R and code/lib/ are as registered (plan/code)." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "The plan puts a table of the per-association results in the paper; at 40 rows it is results/associations.csv instead, linked from the paper, with the three associations that differ significantly from their published estimates in a table in the paper." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this correction, after the results were known and in answer to review: each departure bearing on a headline estimate carries its evidence (paper, data, or unresolved), as code/lib/pipeline.R defines them, and code/lib/summary.R counts the departures and papers by evidence. The lead agent and a second agent, blind to the first, coded each under a written codebook; data/departure_coding.csv holds both codings, and the label in each association file is the settled one. The registered departure records are otherwise unchanged except for the one below." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "code/associations/row311.R recorded as a model departure that the paper's Section 2.3 lists glucose and lipids among the confounders included while Model III, as its Section 2.4 and Table 3 give it and as reproduced, leaves them out. Since the paper describes its model both ways, the departure is now of kind reporting, so the papers with a model departure bearing on the headline fall from five to four." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this correction, in answer to review: the test of each 2021-2023 estimate against the published one and against the harmonized estimate on the paper's cycles are repeated with t distributions on the design degrees of freedom of the fits compared, with the same correction. The registered z tests remain the ones the claims count first." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this correction, in answer to review: each test's power is also computed at the Bonferroni level, 0.05 over the 40 tests, a lower bound on the power of the Benjamini-Hochberg replication decision, as the registered power at 0.05 is an upper bound. The registered definition of an informative test is unchanged." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "The plan says that in 2021-2023 an analysis re-run unweighted is also reported weighted. The first two versions fitted these 15 analyses weighted (replication_weighted in results/associations.json) but reported neither the fits nor the replication count they give. This version reports both: code/lib/pipeline.R corrects the weighted fits' p-values over the 40 associations with the registered fits of the other 25, as the paper-weight sensitivity is corrected, and code/lib/summary.R counts the replications (R1 replication.design.weighted), which C1 states beside the registered count." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, after the results were known and in answer to the second version's reviews: each of the 15 analyses re-run unweighted is also fitted in 2021-2023 with its unweighted estimate and variance from the survey design (every participant weighted equally within the masked strata and PSUs, t on the design degrees of freedom; equal_weights in code/lib/model.R), which separates the variance the unweighted fits understate from the change of estimand that weighting brings. Its replication count is R1 replication.design.design, which C9 states." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, in answer to review: the plan replaces the examination weight with the phlebotomy weight when a blood analyte is in the model, so an analysis of a blood analyte on the interview weight kept it (row284, serum albumin). A sensitivity analysis fits such analyses with the phlebotomy weight (blood_replaces in code/lib/design.R), and C10 states row284's estimate with it." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, in answer to review: each test's power to detect half the published effect, since effects published for crossing a significance threshold are expected to be inflated (power_half; R1 replication.informative_half). The registered definition of an informative test is unchanged." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, in answer to the second version's methods review: each analysis as its paper describes it, on the paper's own cycles. The described version puts back, as the text describes them, every departure bearing on the headline that the re-implementation follows and that is a computation (coding, sample, weighting, or model; a reporting departure records two of the paper's accounts disagreeing, not a computation), all together, reading the text literally and, between two readings, taking the one that changes the followed computation least; a departure whose text version cannot be built from the public files is listed as not applied, with its reason. The rules and the measures (the described estimate outside the published interval over all 40; significance with the published sign among the analyses it changes, as described and as followed; the ratio to the published estimate) were written before any described version but one was fitted (code/described_rules.md), though the coding check had already fitted many single departures' text versions as variants. Each association file's `described` field records what it puts back; code/lib/pipeline.R fits it and refuses a file that leaves a qualifying departure unnamed, and code/lib/summary.R counts the results (R1 reproduction.described)." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, after the results were known and in answer to the second version's domain review (which asked not to attribute the replication rate to power without evidence): the registered replication tests simulated as if every association were true at a given fraction of its published size, each test's statistic drawn from its own 2021-2023 standard error and degrees of freedom, the tests independent, decided as registered (published sign, Benjamini-Hochberg over 40), 10,000 draws with seed 20261019 (expected_replications in code/lib/summary.R; R1 replication.expected). It reports the count expected at the published size and at half of it, the chance of as few replications as observed, and the fraction of the published size at which the observed count is the one expected, with the fractions, in steps of 0.05, that leave it within the central 95% of draws. C15 states them." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added in this version, in answer to reviews noting that the departure records rest on the lead agent's notes: 20 of the 102 departures bearing on a headline estimate, drawn at random from the sorted keys with seed 20261019, were each checked against the paper's full text by an agent that recorded none of them, for whether the record describes the paper correctly (data/departure_validation.csv; code/lib/summary.R checks the file holds exactly the departures drawn and counts the verdicts, R1 reproduction.departures.validation). Statements about what recomputation gives were not part of the check." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "Two departure records the check against the papers found misstating them were corrected, and the check counts both as partly, under its definition (the record misstates the paper but the departure stands), though the checking agent had marked them confirmed with a note: row096's fifth said the mean ferritin of 54.04 ng/ml is Table 1's, where the paper prints it in the Results paragraph that introduces Table 1, and row220's second gave the Methods' definition of heavy drinking without its binge-drinking criterion. Nothing computed changes." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "After a mock review of a draft of this version, in two rounds. First: rows 218 (the weight-adjusted-waist index), 303 (the systemic immune-inflammation index), and 327 (the US fatty liver index) had taken a printed formula that contradicts the named index literally (one fit did not converge), where row062's eGFR had kept the named equation; they were made consistent. Second: applying the second version's precedent for row311 (a departure where the paper describes its model two ways is a reporting departure) to every file, the eight departures where one of the paper's own accounts, or its own steps read literally, gives the computation the paper version follows are now of kind reporting: row062's first, row087's second and third, row131's third, row218's third and fourth, row303's second, and row327's second. They had been in the described versions' applies with no change; the count of papers with a departure of each kind, and of papers whose paper shows a departure of computation, change accordingly. code/described_rules.md, in the bundle, records the rules and when each was set; code/described_checks.md holds the two blind checkers' reports." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "unstated", "detail": "Added after the mock review: the simulation of how many replications the tests could find is repeated with the design-respecting fits of the analyses re-run unweighted (expected_design, expected_weighted in R1); the effect ratio with the registered weighted fits, and over the analyses re-run unweighted and the others apart (ratio_weighted, ratio_unweighted, ratio_others); how the design-based standard errors compare with the unweighted fits' (design.design.se_ratio); the papers whose departure the paper itself shows in a computation rather than only in a number reported (shown_by_paper_computation); how far the described estimates move from the followed ones (described.to_followed_changed, described.within_one_percent); the between-cycle test for each association and the t test of the phlebotomy and weighted refits against the published estimate (R2 heterogeneity_q, difference_q_t_phlebotomy, difference_q_weighted); and, for row284, a variant putting 2017-2018's albumin on 2015-2016's analyzer scale by NCHS's bridging equation (BIOPRO_J: DxC 660i = 1.044 x Cobas 6000 + 0.01128 g/dL), which NCHS recommends when combining the cycles." }, { "from": "prereg:66c14e8df19aafb51c6890eb73fc46addc6eb292998aa04c081f412f2df6e299", "kind": "changed", "detail": "data/departure_validation.csv gains a checker_verdict column, the checking agent's own verdict (confirmed for all 20), beside the verdict counted under the check's definitions, where the lead agent counted two as partly; R1 reproduction.departures.validation.misstating gives the share of records misstating anything about their paper." } ]