Current replication
Spearman rank correlation
Not measuredNo current measurement is available in this registry.
Research study · 2019
Preferences of first-degree relatives of gastric cancer patients for gastric cancer screening: a discrete choice experiment
Explore the published experiment and its available replication evidence.
Read the paper ↗01 / Human finding
What the study found
A reviewed summary is not available in this registry. Consult the published paper for the original finding.
Answer-key review: Not reviewed
02 / Experiment design
Who chose, and what varied
- Country
- Sichuan, Jilin, Gansu, Beijing and Hainan, China
- Human respondents
- Not recorded
- Attributes
- 5
- Task
- discrete choice
03 / Synthetic replication
Registry replication results
Paper-era reference
2023 CLM mean Spearman ρ
Not measuredHistorical results use their original estimator and reference. They are not a matched comparison with the current run.
Detailed replication evidence · 2026-11-03
Right direction on 7 of 8 significant effects, close ranking (ρ = 0.81); sizes not comparable (the paper reports logit coefficients).
Use it to screen which attributes move choice; sizes need the paper's respondent data, and confirm the reversed effect with people.
- direction
- mixed
- rank
- holds
- size
- not comparable
Supported uses
- Ranking attributes and levels by how strongly they move choice.
Limits to keep in view
- Decisions that hinge on the attributes whose direction flipped (listed below).
- Effect sizes in the paper's units; only direction, rank and shape are compared.
Effects on a common scale
Human effects: logit coefficient. Synthetic effects are multiplied by the record’s calibration factor, k = 5.1777. Calibration aligns magnitude; it does not establish agreement in unadjusted effect sizes.
| Attribute / level | Human | Human 95% half-width | Calibrated synthetic |
|---|---|---|---|
| Cost(¥): 200 CNY (reference) | 0.000 | Not measured | 0.000 |
| Cost(¥): 400 CNY | -0.114 | Not measured | -0.332 |
| Cost(¥): 600 CNY | -0.228 | Not measured | -0.304 |
| Waiting time, hour: 5 h (reference) | 0.000 | Not measured | 0.000 |
| Waiting time, hour: 3 h | 0.292 | Not measured | 0.547 |
| Waiting time, hour: 1 h | 0.542 | Not measured | 1.103 |
| Pain: Severe (reference) | 0.000 | Not measured | 0.000 |
| Pain: Mild | 0.707 | Not measured | 0.976 |
| Pain: None | 1.340 | Not measured | 1.508 |
| Frequency: Once every three years (reference) | 0.000 | Not measured | 0.000 |
| Frequency: Once every two years | 0.377 | Not measured | 0.154 |
| Frequency: Once a year | 0.638 | Not measured | -0.019 |
| Sensitivity: 35% (reference) | 0.000 | Not measured | 0.000 |
| Sensitivity: 65% | 0.897 | Not measured | 0.419 |
| Sensitivity: 95% | 1.424 | Not measured | 1.197 |
Subgroups · screen, not a result
Split by sex: 0 of 1 significant differences keep their direction in the simulation (0% to 79%)
screen, not a result. Each row is the female group’s effect minus the male group’s, human in logit, simulated on the simulation’s own scale, so compare direction, not size. A Jev fast-small screen makes no comparison claim.
- Sign agreement
- 0/1, 0%
- 95% interval
- 0% to 79%
- Shuffle null
- 54%, p 0.61
- Shape ρ
- -0.14, null p 0.92
- Scale ratio
- 1.24×
| Group | Simulated n | People n | ρ vs paper |
|---|---|---|---|
| female | 73 | not given | 0.48 |
| male | 77 | not given | 0.40 |
| Attribute / level | People Δ (95%) | Significant | Simulated Δ (own scale) | Direction |
|---|---|---|---|---|
| Waiting time, hour: 3 h | +0.19 ±0.38 | no | +0.05 ±0.07 | not compared |
| Waiting time, hour: 1 h | -0.24 ±0.58 | no | +0.04 ±0.07 | not compared |
| Pain: Mild | +0.00 ±0.44 | no | +0.02 ±0.08 | not compared |
| Pain: None | +1.05 ±0.70 | yes | -0.00 ±0.08 | opposite |
| Frequency: Once every two years | -0.16 ±0.46 | no | +0.07 ±0.08 | not compared |
| Frequency: Once a year | +0.24 ±0.54 | no | +0.07 ±0.08 | not compared |
| Sensitivity: 65% | -0.43 ±0.46 | no | +0.02 ±0.07 | not compared |
| Sensitivity: 95% | +0.10 ±0.62 | no | +0.02 ±0.08 | not compared |
04 / Evidence and limitations
Follow the evidence
Missing results mean unmeasured, not zero agreement.
- missing parent paper url
- missing canonical pdf
- missing historic score
- missing resolved wandb run