OpenPsy

PRICE-02: the price-framing replication with Sol and Terra

In this replication of PRICE-01, GPT-6 Sol and GPT-5.6 Terra each named a higher price under the gain frame. Sol’s observed mean was $24.36 under gain and $14.08 under loss; Terra’s was $18.95 and $9.37. The main collection is closed: 153 of 160 scheduled sessions were complete and eligible. The planned exact tests find higher price and purchase likelihood under gain for each model (all four Holm-adjusted p values < .001). The price interaction is not significant (p = .732). Numerical inference is available below; rationale coding and the final signed study record remain pending.

Numerical inference available · coding and app integration pending

PRICE-01 remains the original Claude Opus 5 / Sonnet 5.5 study. PRICE-02 is the presentation label used here for its Sol/Terra replication. The frozen native campaign retains its original identifiers and records.

What was asked

The scenario and the order of the six participant turns were retained from PRICE-01. A new infection is expected to affect 900 people in a region where the participant lives. A pharmacy will sell Treatment A, a one-season preventive course. Similar courses sell for $25. If all 900 people take it, the gain frame says 300 will be protected; the loss frame says 600 will be left unprotected. These describe the same outcome.

The two-by-two design crosses Frame (gain, loss) with Model (GPT-6 Sol, GPT-5.6 Terra). The primary outcome is the maximum price for one course for oneself, in whole US dollars from 0 to 60. The secondary outcome is likelihood of buying at $25, from 1 (not at all likely) to 7 (very likely). A short price rationale and protected/unprotected count checks complete the instrument.

Two samples each scheduled 20 sessions per condition. The 40-session measurement pilot and four-session execution pilot are excluded from every main-study summary on this page. Complete six-turn sessions are eligible; seven protocol failures remain missing and were not replaced. This whole-session rule is stricter than PRICE-01’s per-outcome availability.

Numerical results

Maximum price in US dollars: Frame labels the rows and Model the columns. Each condition gives its observed mean, sample SD and valid count; marginal summaries pool the observed values. The grand-total corner is blank.
Maximum price in US dollars: Frame labels the rows and Model the columns. Each condition gives its observed mean, sample SD and valid count; marginal summaries pool the observed values. The grand-total corner is blank. Download SVG.
Maximum price (0–60 US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. The dashed line marks the $25 reference.
Maximum price (0–60 US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. The dashed line marks the $25 reference. Download SVG.
Maximum price (0–60 US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. The dashed line marks the $25 reference.
Maximum price (0–60 US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. The dashed line marks the $25 reference. Download SVG.
Maximum price (US dollars)
Frame / ModelGPT-6 SolGPT-5.6 Terra
Gain frameM = 24.36
SD = 2.61, n = 39
95% CI [23.51, 25.21]
1 of 40 missing
M = 18.95
SD = 8.10, n = 37
95% CI [16.25, 21.65]
3 of 40 missing
Loss frameM = 14.08
SD = 7.86, n = 39
95% CI [11.53, 16.63]
1 of 40 missing
M = 9.37
SD = 4.07, n = 38
95% CI [8.03, 10.71]
2 of 40 missing

Gain minus loss is $10.28 for Sol and $9.58 for Terra. Both frame comparisons are significant in the planned exact tests: Holm-adjusted p = 2.700 × 10−9 for Sol and 1.823 × 10−7 for Terra. The sample-blocked price interaction is not significant (p = .732), so the difference between their framing effects is not established.

Price response distribution and exact counts
Counts of observed whole-dollar prices, separated by frame and model. Missing sessions are excluded rather than assigned a zero. Exact counts are available in the response-distribution table below.
Counts of observed whole-dollar prices, separated by frame and model. Missing sessions are excluded rather than assigned a zero. Exact counts are available in the response-distribution table below. Download SVG.

Likelihood to buy at $25

Likelihood to buy at $25 on the 1–7 scale: Frame labels the rows and Model the columns. Each condition gives its observed mean, sample SD and valid count; marginal summaries pool the observed values.
Likelihood to buy at $25 on the 1–7 scale: Frame labels the rows and Model the columns. Each condition gives its observed mean, sample SD and valid count; marginal summaries pool the observed values. Download SVG.
Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending.
Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. Download SVG.
Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending.
Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t intervals for each mean. Pooled bar-chart stars use the exact tests with Holm correction across all four comparisons for that outcome. The line view shows the same means and intervals. Coding and final study recording remain pending. Download SVG.
Likelihood to buy at $25 (1–7)
Frame / ModelGPT-6 SolGPT-5.6 Terra
Gain frameM = 4.90
SD = 0.97, n = 39
95% CI [4.58, 5.21]
1 of 40 missing
M = 4.70
SD = 2.75, n = 37
95% CI [3.79, 5.62]
3 of 40 missing
Loss frameM = 2.56
SD = 1.65, n = 39
95% CI [2.03, 3.10]
1 of 40 missing
M = 1.47
SD = 1.37, n = 38
95% CI [1.02, 1.92]
2 of 40 missing

Sol’s mean likelihood was 4.90 under gain and 2.56 under loss, a difference of 2.33 points. Terra’s means were 4.70 and 1.47, a difference of 3.23 points. Both gain-over-loss comparisons are significant: Holm-adjusted p = 4.923 × 10−9 for Sol and 1.586 × 10−7 for Terra. Sol and Terra do not differ significantly under gain (p = .691), but do under loss (p = .00356). These ratings concern a stated intention in this scenario, not an observed purchase.

Likelihood response distribution and exact counts
Counts at each likelihood rating from 1 to 7, separated by frame and model. All 153 eligible responses are included; seven missing sessions remain outside the distributions.
Counts at each likelihood rating from 1 to 7, separated by frame and model. All 153 eligible responses are included; seven missing sessions remain outside the distributions. Download SVG.

What the rationales say

The five original rationale codes are retained. The final coding package contains 154 captured rationales: 153 from eligible sessions and one for audit only. Final Claude Sonnet 5 coding has not run; there are no code counts, code charts or agreement estimates to report. A pending code is not a zero.

Fresh calibration and both pilot coding stages were completed. The final transfer remains on hold pending explicit approval for the prepared rationale payload. The page and its downloads contain only numerical summaries, not rationale text.

Rationale coding status
CodeGain frame, GPT-6 SolGain frame, GPT-5.6 TerraLoss frame, GPT-6 SolLoss frame, GPT-5.6 Terra
Unprotected share lowers the price (SHORTFALL)PendingPendingPendingPending
Some protection is worth paying for (PROTECTION_VALUE)PendingPendingPendingPending
Framing recognised (FRAME_EQUIVALENCE)PendingPendingPendingPending
$25 reference sets the price (REFERENCE_PRICE)PendingPendingPendingPending
Price derived by arithmetic (CALCULATED_PRICE)PendingPendingPendingPending

Manipulation checks

Every eligible complete session returned 300 protected and 600 unprotected. Both checks were correct in 153 of 153 observed eligible answers. The seven protocol failures remain missing against the scheduled denominator of 160.

Manipulation checks
CheckExpected answerCorrect / observedIncorrectMissingScheduled
Protected count300153 / 15307160
Unprotected count600153 / 15307160
Collection completeness by Frame and Model. Sol gain and loss each have 39 of 40 eligible sessions; Terra gain has 37 of 40 and loss 38 of 40. Seven protocol failures remain missing.
Collection completeness by Frame and Model. Sol gain and loss each have 39 of 40 eligible sessions; Terra gain has 37 of 40 and loss 38 of 40. Seven protocol failures remain missing. Download SVG.

The two samples

The two samples are blocks within the same planned collection. They are separate groups of sessions, not repeated measurements or independent studies. The table and charts retain each sample’s actual count. No sample-specific hypothesis verdict is supplied.

Maximum price (US dollars) by sample
SampleGain frame, GPT-6 SolGain frame, GPT-5.6 TerraLoss frame, GPT-6 SolLoss frame, GPT-5.6 Terra
Sample 1M = 24.74
SD = 1.15, n = 19
M = 18.58
SD = 8.17, n = 19
M = 12.85
SD = 7.25, n = 20
M = 9.37
SD = 4.13, n = 19
Sample 2M = 24.00
SD = 3.48, n = 20
M = 19.33
SD = 8.25, n = 18
M = 15.37
SD = 8.46, n = 19
M = 9.37
SD = 4.13, n = 19
Sample 1: Maximum price (US dollars) — bar and line charts
Sample 1: Maximum price (US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 1: Maximum price (US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 1: Maximum price (US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 1: Maximum price (US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 2: Maximum price (US dollars) — bar and line charts
Sample 2: Maximum price (US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 2: Maximum price (US dollars): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 2: Maximum price (US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 2: Maximum price (US dollars): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Likelihood to buy at $25 (1–7) by sample
SampleGain frame, GPT-6 SolGain frame, GPT-5.6 TerraLoss frame, GPT-6 SolLoss frame, GPT-5.6 Terra
Sample 1M = 5.16
SD = 0.76, n = 19
M = 4.47
SD = 2.78, n = 19
M = 2.25
SD = 1.59, n = 20
M = 1.58
SD = 1.39, n = 19
Sample 2M = 4.65
SD = 1.09, n = 20
M = 4.94
SD = 2.78, n = 18
M = 2.89
SD = 1.70, n = 19
M = 1.37
SD = 1.38, n = 19
Sample 1: Likelihood to buy at $25 (1–7) — bar and line charts
Sample 1: Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 1: Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 1: Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 1: Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 2: Likelihood to buy at $25 (1–7) — bar and line charts
Sample 2: Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 2: Likelihood to buy at $25 (1–7): observed means by Frame and Model. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.
Sample 2: Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range.
Sample 2: Likelihood to buy at $25 (1–7): the same means as lines across the two frame conditions. These are categorical comparisons, not changes over time. Error bars are individual unadjusted 95% Student t mean intervals. These sample plots have no significance marks: the registered tests pool the samples while accounting for their blocks. Axes show the full intervals, including any portion outside the response range. Download SVG.

Inferential statistics

These are the frozen plan’s numerical tests on 153 eligible sessions. All four cells meet the 90% valid/scheduled threshold, and all 160 scheduled sessions are accounted for. The code outcomes and final signed study record remain pending; these tables are a partial numerical analysis.

Each outcome has four two-sided exact permutation tests, permuting within the two sample blocks. Their statistic is the absolute sum of the two within-sample mean differences. The displayed pooled mean difference is a descriptive effect size and need not equal half that statistic when counts differ. Holm adjustment includes all four comparisons within each outcome, separately for price and likelihood. No comparison was dropped from either family.

Maximum price: four exact sample-stratified comparisons
ComparisonFirst nSecond nPooled mean differenceSigned sum of sample differencesRaw pHolm p
Gain − loss, GPT-6 Sol393910.28220.5186.75e-102.7e-09
Gain − loss, GPT-5.6 Terra37389.57819.1756.08e-081.82e-07
Sol − Terra, gain frame39375.41310.8250.0001550.00031
Sol − Terra, loss frame39384.7099.4820.0020.002
Likelihood to buy: separate four-comparison family
ComparisonFirst nSecond nPooled mean differenceSigned sum of sample differencesRaw pHolm p
Gain − loss, GPT-6 Sol39392.3334.6631.23e-094.92e-09
Gain − loss, GPT-5.6 Terra37383.2296.4715.29e-081.59e-07
Sol − Terra, gain frame39370.1950.3900.6910.691
Sol − Terra, loss frame39381.0902.1970.0020.004

Differences are in US dollars for price and rating points for likelihood. The exact test compares absolute values of the signed sample sum; signs are retained here to make direction readable. The permutation method does not supply a conventional standard error, t statistic, degrees of freedom or contrast confidence interval, so none is invented. Full-precision p values and exact allocation totals are in the numerical JSON download.

Maximum price: sample-blocked linear model
Termb (US dollars)SEtdfp95% CI lower95% CI upper
main effect of Frame9.9260.99210.0071482.52e-187.96611.886
main effect of Model5.0560.9925.0981481.04e-063.0967.017
interaction0.6801.9840.3431480.732-3.2414.601

The declared price model includes a sample intercept, Frame, Model and their interaction. Gain and Sol are coded +0.5; loss and Terra −0.5. The three Wald t tests use 148 residual degrees of freedom. Their p values are unadjusted, as declared. The coefficient intervals are supplementary, unadjusted 95% Wald t intervals.

The price interaction is b = 0.680, t(148) = 0.343, p = .732. This provides no significant evidence that the two models’ price-framing effects differ; it does not establish equivalence. No omnibus model for likelihood was registered, and none is added here.

Independent arithmetic verification

Independent enumeration reproduced all 32 observed and sensitivity exact-test probabilities and all eight Holm families. An independent high-precision calculation found a small approximation error in the unchanged shared Student t routine: at most 8.07 × 10−9 relative difference in the model p values and 3.59 × 10−9 absolute difference in coefficient interval endpoints. This changes neither the displayed values nor any .05 decision. The frozen calculation is retained.

Numerical decision-rule results; final study recording pending
IDModelOutcomePredictionHolm pNumerical result
H1aGPT-5.6 TerraMaximum priceGain > loss1.82e-07Criterion met
H1bGPT-6 SolMaximum priceGain > loss2.7e-09Criterion met
H2aGPT-5.6 TerraLikelihood to buyGain > loss1.59e-07Criterion met
H2bGPT-6 SolLikelihood to buyGain > loss4.92e-09Criterion met

All four numerical hypotheses meet their frozen criterion: a significant effect in the predicted direction at Holm-adjusted α = .05. This is the mechanical numerical result, not a claim that the full study has been finalized or admitted into the OpenPsy app. The original Opus-versus-Sonnet interaction hypothesis H1c was not transferred to these models.

All pooled and sample-specific means with 95% confidence intervals
Cell means and individual 95% confidence intervals
OutcomeSampleModelFramenMeanSD95% CI lower95% CI upper
Maximum price (USD)PooledGPT-6 SolGain frame3924.3592.61123.51325.205
Maximum price (USD)PooledGPT-5.6 TerraGain frame3718.9468.10016.24521.647
Maximum price (USD)PooledGPT-6 SolLoss frame3914.0777.86211.52816.625
Maximum price (USD)PooledGPT-5.6 TerraLoss frame389.3684.0708.03110.706
Maximum price (USD)Sample 1GPT-6 SolGain frame1924.7371.14724.18425.290
Maximum price (USD)Sample 1GPT-5.6 TerraGain frame1918.5798.16714.64322.515
Maximum price (USD)Sample 1GPT-6 SolLoss frame2012.8507.2509.45716.243
Maximum price (USD)Sample 1GPT-5.6 TerraLoss frame199.3684.1267.38011.357
Maximum price (USD)Sample 2GPT-6 SolGain frame20243.47922.37225.628
Maximum price (USD)Sample 2GPT-5.6 TerraGain frame1819.3338.24615.23323.434
Maximum price (USD)Sample 2GPT-6 SolLoss frame1915.3688.46011.29119.446
Maximum price (USD)Sample 2GPT-5.6 TerraLoss frame199.3684.1267.38011.357
Likelihood (1–7)PooledGPT-6 SolGain frame394.8970.9684.5845.211
Likelihood (1–7)PooledGPT-5.6 TerraGain frame374.7032.7473.7875.619
Likelihood (1–7)PooledGPT-6 SolLoss frame392.5641.6512.0293.099
Likelihood (1–7)PooledGPT-5.6 TerraLoss frame381.4741.3701.0231.924
Likelihood (1–7)Sample 1GPT-6 SolGain frame195.1580.7654.7895.526
Likelihood (1–7)Sample 1GPT-5.6 TerraGain frame194.4742.7763.1365.812
Likelihood (1–7)Sample 1GPT-6 SolLoss frame202.2501.5851.5082.992
Likelihood (1–7)Sample 1GPT-5.6 TerraLoss frame191.5791.3870.9102.247
Likelihood (1–7)Sample 2GPT-6 SolGain frame204.6501.0894.1405.160
Likelihood (1–7)Sample 2GPT-5.6 TerraGain frame184.9442.7753.5646.325
Likelihood (1–7)Sample 2GPT-6 SolLoss frame192.8951.6962.0773.712
Likelihood (1–7)Sample 2GPT-5.6 TerraLoss frame191.3681.3830.7022.035

Each interval is mean ± t(0.975, n − 1) × SD/√n. These are individual unadjusted mean intervals, not simultaneous intervals across conditions. The same calculation is used in OpenPsy’s mean figures. Interval overlap is not the significance test. A t interval may extend beyond the permitted response range; plots show its full extent rather than clipping it.

Missing-data sensitivity

The original grammar-only sensitivity rule covers invalid, refused and provider-truncated outcomes. None of the eligible numerical answers had those classifications. Its bounds and four decision-rule results therefore equal the observed analysis. Native whole-session failures are a different classification and must not be relabeled as provider truncation.

The separate native sensitivity includes all seven unavailable scheduled outcomes. It evaluates all missing values at the scale minimum, all at the maximum, and the assignment against the gain-over-loss prediction: missing gain outcomes at the minimum and missing loss outcomes at the maximum. Every scenario reruns both complete four-comparison Holm families. These are endpoint assumptions, not imputed observations or confidence intervals.

Native missingness: all four numerical criteria under each scenario
ScenarioIDModelOutcomeHolm pNumerical result
All unavailable at minimumH1aGPT-5.6 TerraMaximum price7.15e-06Criterion met
All unavailable at minimumH1bGPT-6 SolMaximum price3.6e-08Criterion met
All unavailable at minimumH2aGPT-5.6 TerraLikelihood to buy7.3e-07Criterion met
All unavailable at minimumH2bGPT-6 SolLikelihood to buy2.15e-08Criterion met
All unavailable at maximumH1aGPT-5.6 TerraMaximum price0.000972Criterion met
All unavailable at maximumH1bGPT-6 SolMaximum price1.6e-06Criterion met
All unavailable at maximumH2aGPT-5.6 TerraLikelihood to buy9.22e-07Criterion met
All unavailable at maximumH2bGPT-6 SolLikelihood to buy2.62e-08Criterion met
Against the gain > loss predictionH1aGPT-5.6 TerraMaximum price0.039Criterion met
Against the gain > loss predictionH1bGPT-6 SolMaximum price2.78e-05Criterion met
Against the gain > loss predictionH2aGPT-5.6 TerraLikelihood to buy2.51e-05Criterion met
Against the gain > loss predictionH2bGPT-6 SolLikelihood to buy2.92e-07Criterion met

All four criteria remain met in these endpoint scenarios. The least-favourable Terra price comparison is closest to the threshold, with Holm-adjusted p = .039. These bounds address the seven missing numerical outcomes under the stated range assumptions; they do not address route differences, unverified provider controls or other sources of bias.

Native all-unavailable bounds on each cell mean
OutcomeModelFrameObserved nMissingObserved meanLower boundUpper bound
Maximum price (USD)GPT-6 SolGain frame39124.35923.75025.250
Maximum price (USD)GPT-5.6 TerraGain frame37318.94617.52522.025
Maximum price (USD)GPT-6 SolLoss frame39114.07713.72515.225
Maximum price (USD)GPT-5.6 TerraLoss frame3829.3688.90011.900
Likelihood (1–7)GPT-6 SolGain frame3914.8974.8004.950
Likelihood (1–7)GPT-5.6 TerraGain frame3734.7034.4254.875
Likelihood (1–7)GPT-6 SolLoss frame3912.5642.5252.675
Likelihood (1–7)GPT-5.6 TerraLoss frame3821.4741.4501.750
Every sensitivity comparison and price-model test
All 24 exact comparisons in the three native endpoint scenarios
ScenarioOutcomeComparisonFirst nSecond nPooled mean differenceSigned sum of sample differencesRaw pHolm p
all-minimumMaximum price (USD)Gain − loss, GPT-6 Sol404010.02520.0509e-093.6e-08
all-minimumMaximum price (USD)Gain − loss, GPT-5.6 Terra40408.62517.2502.38e-067.15e-06
all-minimumMaximum price (USD)Sol − Terra, gain frame40406.22512.4500.0002740.000547
all-minimumMaximum price (USD)Sol − Terra, loss frame40404.8259.6500.0010.001
all-minimumLikelihood (1–7)Gain − loss, GPT-6 Sol40402.2754.5505.38e-092.15e-08
all-minimumLikelihood (1–7)Gain − loss, GPT-5.6 Terra40402.9755.9502.43e-077.3e-07
all-minimumLikelihood (1–7)Sol − Terra, gain frame40400.3750.7500.4700.470
all-minimumLikelihood (1–7)Sol − Terra, loss frame40401.0752.1500.0020.005
all-maximumMaximum price (USD)Gain − loss, GPT-6 Sol404010.02520.0504e-071.6e-06
all-maximumMaximum price (USD)Gain − loss, GPT-5.6 Terra404010.12520.2500.0003240.000972
all-maximumMaximum price (USD)Sol − Terra, gain frame40403.2256.4500.1910.382
all-maximumMaximum price (USD)Sol − Terra, loss frame40403.3256.6500.2040.382
all-maximumLikelihood (1–7)Gain − loss, GPT-6 Sol40402.2754.5506.54e-092.62e-08
all-maximumLikelihood (1–7)Gain − loss, GPT-5.6 Terra40403.1256.2503.07e-079.22e-07
all-maximumLikelihood (1–7)Sol − Terra, gain frame40400.0750.1500.9140.914
all-maximumLikelihood (1–7)Sol − Terra, loss frame40400.9251.8500.0280.056
gain-minimum-loss-maximumMaximum price (USD)Gain − loss, GPT-6 Sol40408.52517.0506.95e-062.78e-05
gain-minimum-loss-maximumMaximum price (USD)Gain − loss, GPT-5.6 Terra40405.62511.2500.0200.039
gain-minimum-loss-maximumMaximum price (USD)Sol − Terra, gain frame40406.22512.4500.0002740.000821
gain-minimum-loss-maximumMaximum price (USD)Sol − Terra, loss frame40403.3256.6500.2040.204
gain-minimum-loss-maximumLikelihood (1–7)Gain − loss, GPT-6 Sol40402.1254.2507.31e-082.92e-07
gain-minimum-loss-maximumLikelihood (1–7)Gain − loss, GPT-5.6 Terra40402.6755.3508.36e-062.51e-05
gain-minimum-loss-maximumLikelihood (1–7)Sol − Terra, gain frame40400.3750.7500.4700.470
gain-minimum-loss-maximumLikelihood (1–7)Sol − Terra, loss frame40400.9251.8500.0280.056
All nine price-model terms in the three native endpoint scenarios
ScenarioTermbSEtdfp
all-minimummain effect of Frame9.3251.1008.4741551.72e-14
all-minimummain effect of Model5.5251.1005.0211551.4e-06
all-minimuminteraction1.4002.2010.6361550.526
all-maximummain effect of Frame10.0751.7215.8541552.77e-08
all-maximummain effect of Model3.2751.7211.9031550.059
all-maximuminteraction-0.1003.442-0.0291550.977
gain-minimum-loss-maximummain effect of Frame7.0751.5044.7051555.59e-06
gain-minimum-loss-maximummain effect of Model4.7751.5043.1751550.002
gain-minimum-loss-maximuminteraction2.9003.0080.9641550.336

What these results support

The four retained gain-over-loss hypotheses meet their numerical decision rule for both maximum price and likelihood to buy. That statement is restricted to the collected sessions, wording and native route. The price interaction is not significant; a non-significant interaction does not establish equal effects.

The missing-data endpoint scenarios retain all four numerical criteria, although the least-favourable Terra price comparison is close to the threshold at Holm-adjusted p = .039. No rationale-based explanation is supplied until the planned coding is complete.

The native campaign remains outside the completed standard Library/app workflow. This partial numerical report does not substitute for completed coding, a fresh signed independent analysis review or final study recording.

How the run went

The 40-session measurement pilot and four-session execution pilot completed, with fresh calibration, collection review and coding. The fixed main schedule then processed all 160 sessions. There are 153 eligible complete sessions and seven known protocol failures, with no unknown or unattempted main-study sessions and no replacements. Protocol rejection is not by itself evidence of a scientific model refusal.

This was a source-pinned native successor with fresh plan, execution and independent-review records. It was not the unchanged original OpenPsy app import-and-run path. The differences below remain part of its interpretation and cannot be repaired by renaming or reformatting the page.

Documented differences from PRICE-01
AreaNative replication record
App and LibraryThe preserved source was checked with the original compiler and canonical Flow simulator. The current Library compiler rejected the historical executableVersion field. This campaign was not imported through the current Library UI.
Participant routeSeparate native Codex execution replaced the Claude Code route. Requested and configured model names were observed; the serving model snapshot was not independently verified.
Reply limitThe original 1,024-token cap could not be sent as a cap parameter through this native API. No extra length instruction was inserted. Observed usage does not verify provider truncation or an enforced limit.
Reasoning, randomness and contextLow reasoning effort was requested. Effective compute, temperature, top-p, account isolation and complete upstream context were not independently verified. Observed tool activity caused protocol refusal.
Execution and eligibilitySessions ran sequentially in the committed randomized order. Only complete six-turn sessions were eligible; no missing main-study session was replaced. PRICE-01 used concurrency four and per-outcome availability.
Administrative second attemptA separately reviewed pre-response startup attempt retained one unknown claimed row and 39 unattempted rows, with zero application participant requests. One prospective administrative successor was approved. This is a substantive exception to unchanged no-repeat compliance.
Pilot-result exposurePilot cell means and contrasts, then four pilot rationale texts, were viewed before confirmatory finalization and disclosed. The pre-participant scientific commitments remained unchanged; reviewer blinding is not claimed.
Pilot gates and powerThe observable native completion endpoint replaced the unavailable provider-truncation gate. Original planning distributions were proxies; target-model power was not validated.
Coding and analysisFresh single-coder calibration and both pilots passed their recorded gates. Scalar inference and both numerical sensitivity analyses have been computed under the frozen rules. Final code analyses, signed study recording and app result integration remain incomplete.

How this relates to PRICE-01

PRICE-01 is the original Claude Opus 5 / Claude Sonnet 5.5 study, run on 6 October 2026. It has a saved OpenPsy study, recorded analysis and a published results page. Its price outcome used 157 available responses, its likelihood outcome 156, and its coded rationales 154; those denominators are specific to the original study.

PRICE-02 is the review label for the GPT-6 Sol / GPT-5.6 Terra replication. The retained campaign identifier is PRICE01-SOL-TERRA-20261007. Historical paths, signed records and source hashes continue to use that original identifier. This presentation mapping does not create a new registered app study or retroactively change collection.

Read the two studies separately: participant models, route controls, context, concurrency and eligibility differ. A comparison between the pages is descriptive and does not isolate a model or provider effect.

Tables and downloads

Download all PRICE-02 figures and tables · Verified descriptive data (JSON)

Full-precision numerical inference (JSON)

Condition summaries — full table
Condition summaries
ModelFrameOutcomeUnitScheduledAnalysedMissingMeanSample SD
GPT-6 SolGain frameMaximum priceUS dollars4039124.362.61
GPT-6 SolGain frameLikelihood to buy1 to 7403914.900.97
GPT-6 SolLoss frameMaximum priceUS dollars4039114.087.86
GPT-6 SolLoss frameLikelihood to buy1 to 7403912.561.65
GPT-5.6 TerraGain frameMaximum priceUS dollars4037318.958.10
GPT-5.6 TerraGain frameLikelihood to buy1 to 7403734.702.75
GPT-5.6 TerraLoss frameMaximum priceUS dollars403829.374.07
GPT-5.6 TerraLoss frameLikelihood to buy1 to 7403821.471.37
Sample summaries — full table
Sample summaries
SampleModelFrameOutcomeUnitScheduledAnalysedMissingMeanSample SD
1GPT-6 SolGain frameMaximum priceUS dollars2019124.741.15
1GPT-6 SolGain frameLikelihood to buy1 to 7201915.160.76
1GPT-6 SolLoss frameMaximum priceUS dollars2020012.857.25
1GPT-6 SolLoss frameLikelihood to buy1 to 7202002.251.59
1GPT-5.6 TerraGain frameMaximum priceUS dollars2019118.588.17
1GPT-5.6 TerraGain frameLikelihood to buy1 to 7201914.472.78
1GPT-5.6 TerraLoss frameMaximum priceUS dollars201919.374.13
1GPT-5.6 TerraLoss frameLikelihood to buy1 to 7201911.581.39
2GPT-6 SolGain frameMaximum priceUS dollars20200243.48
2GPT-6 SolGain frameLikelihood to buy1 to 7202004.651.09
2GPT-6 SolLoss frameMaximum priceUS dollars2019115.378.46
2GPT-6 SolLoss frameLikelihood to buy1 to 7201912.891.70
2GPT-5.6 TerraGain frameMaximum priceUS dollars2018219.338.25
2GPT-5.6 TerraGain frameLikelihood to buy1 to 7201824.942.78
2GPT-5.6 TerraLoss frameMaximum priceUS dollars201919.374.13
2GPT-5.6 TerraLoss frameLikelihood to buy1 to 7201911.371.38
Gain minus loss differences — full table
Gain minus loss differences
SampleModelOutcomeUnitGain NLoss NDifference
PooledGPT-6 SolMaximum priceUS dollars393910.28
PooledGPT-6 SolLikelihood to buyscale points39392.33
PooledGPT-5.6 TerraMaximum priceUS dollars37389.58
PooledGPT-5.6 TerraLikelihood to buyscale points37383.23
1GPT-6 SolMaximum priceUS dollars192011.89
1GPT-6 SolLikelihood to buyscale points19202.91
1GPT-5.6 TerraMaximum priceUS dollars19199.21
1GPT-5.6 TerraLikelihood to buyscale points19192.89
2GPT-6 SolMaximum priceUS dollars20198.63
2GPT-6 SolLikelihood to buyscale points20191.76
2GPT-5.6 TerraMaximum priceUS dollars18199.96
2GPT-5.6 TerraLikelihood to buyscale points18193.58
Observed response distributions — full table
Observed response distributions
ModelFrameOutcomeValueCount
GPT-6 SolGain frameMaximum price00
GPT-6 SolGain frameMaximum price10
GPT-6 SolGain frameMaximum price20
GPT-6 SolGain frameMaximum price30
GPT-6 SolGain frameMaximum price40
GPT-6 SolGain frameMaximum price50
GPT-6 SolGain frameMaximum price60
GPT-6 SolGain frameMaximum price70
GPT-6 SolGain frameMaximum price80
GPT-6 SolGain frameMaximum price90
GPT-6 SolGain frameMaximum price101
GPT-6 SolGain frameMaximum price110
GPT-6 SolGain frameMaximum price120
GPT-6 SolGain frameMaximum price130
GPT-6 SolGain frameMaximum price140
GPT-6 SolGain frameMaximum price150
GPT-6 SolGain frameMaximum price160
GPT-6 SolGain frameMaximum price170
GPT-6 SolGain frameMaximum price180
GPT-6 SolGain frameMaximum price190
GPT-6 SolGain frameMaximum price202
GPT-6 SolGain frameMaximum price210
GPT-6 SolGain frameMaximum price220
GPT-6 SolGain frameMaximum price230
GPT-6 SolGain frameMaximum price240
GPT-6 SolGain frameMaximum price2536
GPT-6 SolGain frameMaximum price260
GPT-6 SolGain frameMaximum price270
GPT-6 SolGain frameMaximum price280
GPT-6 SolGain frameMaximum price290
GPT-6 SolGain frameMaximum price300
GPT-6 SolGain frameMaximum price310
GPT-6 SolGain frameMaximum price320
GPT-6 SolGain frameMaximum price330
GPT-6 SolGain frameMaximum price340
GPT-6 SolGain frameMaximum price350
GPT-6 SolGain frameMaximum price360
GPT-6 SolGain frameMaximum price370
GPT-6 SolGain frameMaximum price380
GPT-6 SolGain frameMaximum price390
GPT-6 SolGain frameMaximum price400
GPT-6 SolGain frameMaximum price410
GPT-6 SolGain frameMaximum price420
GPT-6 SolGain frameMaximum price430
GPT-6 SolGain frameMaximum price440
GPT-6 SolGain frameMaximum price450
GPT-6 SolGain frameMaximum price460
GPT-6 SolGain frameMaximum price470
GPT-6 SolGain frameMaximum price480
GPT-6 SolGain frameMaximum price490
GPT-6 SolGain frameMaximum price500
GPT-6 SolGain frameMaximum price510
GPT-6 SolGain frameMaximum price520
GPT-6 SolGain frameMaximum price530
GPT-6 SolGain frameMaximum price540
GPT-6 SolGain frameMaximum price550
GPT-6 SolGain frameMaximum price560
GPT-6 SolGain frameMaximum price570
GPT-6 SolGain frameMaximum price580
GPT-6 SolGain frameMaximum price590
GPT-6 SolGain frameMaximum price600
GPT-6 SolGain frameLikelihood to buy11
GPT-6 SolGain frameLikelihood to buy20
GPT-6 SolGain frameLikelihood to buy32
GPT-6 SolGain frameLikelihood to buy44
GPT-6 SolGain frameLikelihood to buy524
GPT-6 SolGain frameLikelihood to buy68
GPT-6 SolGain frameLikelihood to buy70
GPT-6 SolLoss frameMaximum price00
GPT-6 SolLoss frameMaximum price10
GPT-6 SolLoss frameMaximum price20
GPT-6 SolLoss frameMaximum price30
GPT-6 SolLoss frameMaximum price40
GPT-6 SolLoss frameMaximum price50
GPT-6 SolLoss frameMaximum price60
GPT-6 SolLoss frameMaximum price70
GPT-6 SolLoss frameMaximum price818
GPT-6 SolLoss frameMaximum price90
GPT-6 SolLoss frameMaximum price108
GPT-6 SolLoss frameMaximum price110
GPT-6 SolLoss frameMaximum price120
GPT-6 SolLoss frameMaximum price130
GPT-6 SolLoss frameMaximum price140
GPT-6 SolLoss frameMaximum price150
GPT-6 SolLoss frameMaximum price160
GPT-6 SolLoss frameMaximum price170
GPT-6 SolLoss frameMaximum price180
GPT-6 SolLoss frameMaximum price190
GPT-6 SolLoss frameMaximum price200
GPT-6 SolLoss frameMaximum price210
GPT-6 SolLoss frameMaximum price220
GPT-6 SolLoss frameMaximum price230
GPT-6 SolLoss frameMaximum price240
GPT-6 SolLoss frameMaximum price2513
GPT-6 SolLoss frameMaximum price260
GPT-6 SolLoss frameMaximum price270
GPT-6 SolLoss frameMaximum price280
GPT-6 SolLoss frameMaximum price290
GPT-6 SolLoss frameMaximum price300
GPT-6 SolLoss frameMaximum price310
GPT-6 SolLoss frameMaximum price320
GPT-6 SolLoss frameMaximum price330
GPT-6 SolLoss frameMaximum price340
GPT-6 SolLoss frameMaximum price350
GPT-6 SolLoss frameMaximum price360
GPT-6 SolLoss frameMaximum price370
GPT-6 SolLoss frameMaximum price380
GPT-6 SolLoss frameMaximum price390
GPT-6 SolLoss frameMaximum price400
GPT-6 SolLoss frameMaximum price410
GPT-6 SolLoss frameMaximum price420
GPT-6 SolLoss frameMaximum price430
GPT-6 SolLoss frameMaximum price440
GPT-6 SolLoss frameMaximum price450
GPT-6 SolLoss frameMaximum price460
GPT-6 SolLoss frameMaximum price470
GPT-6 SolLoss frameMaximum price480
GPT-6 SolLoss frameMaximum price490
GPT-6 SolLoss frameMaximum price500
GPT-6 SolLoss frameMaximum price510
GPT-6 SolLoss frameMaximum price520
GPT-6 SolLoss frameMaximum price530
GPT-6 SolLoss frameMaximum price540
GPT-6 SolLoss frameMaximum price550
GPT-6 SolLoss frameMaximum price560
GPT-6 SolLoss frameMaximum price570
GPT-6 SolLoss frameMaximum price580
GPT-6 SolLoss frameMaximum price590
GPT-6 SolLoss frameMaximum price600
GPT-6 SolLoss frameLikelihood to buy114
GPT-6 SolLoss frameLikelihood to buy212
GPT-6 SolLoss frameLikelihood to buy30
GPT-6 SolLoss frameLikelihood to buy43
GPT-6 SolLoss frameLikelihood to buy510
GPT-6 SolLoss frameLikelihood to buy60
GPT-6 SolLoss frameLikelihood to buy70
GPT-5.6 TerraGain frameMaximum price00
GPT-5.6 TerraGain frameMaximum price10
GPT-5.6 TerraGain frameMaximum price20
GPT-5.6 TerraGain frameMaximum price30
GPT-5.6 TerraGain frameMaximum price40
GPT-5.6 TerraGain frameMaximum price50
GPT-5.6 TerraGain frameMaximum price60
GPT-5.6 TerraGain frameMaximum price70
GPT-5.6 TerraGain frameMaximum price812
GPT-5.6 TerraGain frameMaximum price90
GPT-5.6 TerraGain frameMaximum price101
GPT-5.6 TerraGain frameMaximum price110
GPT-5.6 TerraGain frameMaximum price120
GPT-5.6 TerraGain frameMaximum price130
GPT-5.6 TerraGain frameMaximum price140
GPT-5.6 TerraGain frameMaximum price150
GPT-5.6 TerraGain frameMaximum price160
GPT-5.6 TerraGain frameMaximum price170
GPT-5.6 TerraGain frameMaximum price180
GPT-5.6 TerraGain frameMaximum price190
GPT-5.6 TerraGain frameMaximum price201
GPT-5.6 TerraGain frameMaximum price210
GPT-5.6 TerraGain frameMaximum price220
GPT-5.6 TerraGain frameMaximum price230
GPT-5.6 TerraGain frameMaximum price240
GPT-5.6 TerraGain frameMaximum price2523
GPT-5.6 TerraGain frameMaximum price260
GPT-5.6 TerraGain frameMaximum price270
GPT-5.6 TerraGain frameMaximum price280
GPT-5.6 TerraGain frameMaximum price290
GPT-5.6 TerraGain frameMaximum price300
GPT-5.6 TerraGain frameMaximum price310
GPT-5.6 TerraGain frameMaximum price320
GPT-5.6 TerraGain frameMaximum price330
GPT-5.6 TerraGain frameMaximum price340
GPT-5.6 TerraGain frameMaximum price350
GPT-5.6 TerraGain frameMaximum price360
GPT-5.6 TerraGain frameMaximum price370
GPT-5.6 TerraGain frameMaximum price380
GPT-5.6 TerraGain frameMaximum price390
GPT-5.6 TerraGain frameMaximum price400
GPT-5.6 TerraGain frameMaximum price410
GPT-5.6 TerraGain frameMaximum price420
GPT-5.6 TerraGain frameMaximum price430
GPT-5.6 TerraGain frameMaximum price440
GPT-5.6 TerraGain frameMaximum price450
GPT-5.6 TerraGain frameMaximum price460
GPT-5.6 TerraGain frameMaximum price470
GPT-5.6 TerraGain frameMaximum price480
GPT-5.6 TerraGain frameMaximum price490
GPT-5.6 TerraGain frameMaximum price500
GPT-5.6 TerraGain frameMaximum price510
GPT-5.6 TerraGain frameMaximum price520
GPT-5.6 TerraGain frameMaximum price530
GPT-5.6 TerraGain frameMaximum price540
GPT-5.6 TerraGain frameMaximum price550
GPT-5.6 TerraGain frameMaximum price560
GPT-5.6 TerraGain frameMaximum price570
GPT-5.6 TerraGain frameMaximum price580
GPT-5.6 TerraGain frameMaximum price590
GPT-5.6 TerraGain frameMaximum price600
GPT-5.6 TerraGain frameLikelihood to buy111
GPT-5.6 TerraGain frameLikelihood to buy22
GPT-5.6 TerraGain frameLikelihood to buy31
GPT-5.6 TerraGain frameLikelihood to buy40
GPT-5.6 TerraGain frameLikelihood to buy50
GPT-5.6 TerraGain frameLikelihood to buy65
GPT-5.6 TerraGain frameLikelihood to buy718
GPT-5.6 TerraLoss frameMaximum price00
GPT-5.6 TerraLoss frameMaximum price10
GPT-5.6 TerraLoss frameMaximum price20
GPT-5.6 TerraLoss frameMaximum price30
GPT-5.6 TerraLoss frameMaximum price40
GPT-5.6 TerraLoss frameMaximum price50
GPT-5.6 TerraLoss frameMaximum price60
GPT-5.6 TerraLoss frameMaximum price70
GPT-5.6 TerraLoss frameMaximum price832
GPT-5.6 TerraLoss frameMaximum price90
GPT-5.6 TerraLoss frameMaximum price102
GPT-5.6 TerraLoss frameMaximum price110
GPT-5.6 TerraLoss frameMaximum price120
GPT-5.6 TerraLoss frameMaximum price130
GPT-5.6 TerraLoss frameMaximum price140
GPT-5.6 TerraLoss frameMaximum price152
GPT-5.6 TerraLoss frameMaximum price160
GPT-5.6 TerraLoss frameMaximum price170
GPT-5.6 TerraLoss frameMaximum price180
GPT-5.6 TerraLoss frameMaximum price190
GPT-5.6 TerraLoss frameMaximum price200
GPT-5.6 TerraLoss frameMaximum price210
GPT-5.6 TerraLoss frameMaximum price220
GPT-5.6 TerraLoss frameMaximum price230
GPT-5.6 TerraLoss frameMaximum price240
GPT-5.6 TerraLoss frameMaximum price252
GPT-5.6 TerraLoss frameMaximum price260
GPT-5.6 TerraLoss frameMaximum price270
GPT-5.6 TerraLoss frameMaximum price280
GPT-5.6 TerraLoss frameMaximum price290
GPT-5.6 TerraLoss frameMaximum price300
GPT-5.6 TerraLoss frameMaximum price310
GPT-5.6 TerraLoss frameMaximum price320
GPT-5.6 TerraLoss frameMaximum price330
GPT-5.6 TerraLoss frameMaximum price340
GPT-5.6 TerraLoss frameMaximum price350
GPT-5.6 TerraLoss frameMaximum price360
GPT-5.6 TerraLoss frameMaximum price370
GPT-5.6 TerraLoss frameMaximum price380
GPT-5.6 TerraLoss frameMaximum price390
GPT-5.6 TerraLoss frameMaximum price400
GPT-5.6 TerraLoss frameMaximum price410
GPT-5.6 TerraLoss frameMaximum price420
GPT-5.6 TerraLoss frameMaximum price430
GPT-5.6 TerraLoss frameMaximum price440
GPT-5.6 TerraLoss frameMaximum price450
GPT-5.6 TerraLoss frameMaximum price460
GPT-5.6 TerraLoss frameMaximum price470
GPT-5.6 TerraLoss frameMaximum price480
GPT-5.6 TerraLoss frameMaximum price490
GPT-5.6 TerraLoss frameMaximum price500
GPT-5.6 TerraLoss frameMaximum price510
GPT-5.6 TerraLoss frameMaximum price520
GPT-5.6 TerraLoss frameMaximum price530
GPT-5.6 TerraLoss frameMaximum price540
GPT-5.6 TerraLoss frameMaximum price550
GPT-5.6 TerraLoss frameMaximum price560
GPT-5.6 TerraLoss frameMaximum price570
GPT-5.6 TerraLoss frameMaximum price580
GPT-5.6 TerraLoss frameMaximum price590
GPT-5.6 TerraLoss frameMaximum price600
GPT-5.6 TerraLoss frameLikelihood to buy130
GPT-5.6 TerraLoss frameLikelihood to buy26
GPT-5.6 TerraLoss frameLikelihood to buy30
GPT-5.6 TerraLoss frameLikelihood to buy40
GPT-5.6 TerraLoss frameLikelihood to buy50
GPT-5.6 TerraLoss frameLikelihood to buy60
GPT-5.6 TerraLoss frameLikelihood to buy72
Manipulation checks — full table
Manipulation checks
CheckExpectedScheduledValidCorrectIncorrectMissing
Protected count30016015315307
Unprotected count60016015315307

Review and application status

This page presents the partial numerical results for review on OpenPsy. PRICE-01 remains the separate original study. Rationale coding, final signed recording and standard app integration remain pending.

The corrected page reuses the established site’s typography, section structure, two-by-two tables, model colours, figure grammar and accessible bar/line controls. It includes the planned scalar tests and their uncertainty/sensitivity tables. It does not claim that the native study has completed coding, final recording or the canonical app workflow.

What exists and what remains
ItemStatus
Original PRICE-01Saved OpenPsy app study and published recorded-results page
PRICE-02 collectionClosed: 153 complete eligible sessions from 160 scheduled
PRICE-02 Library/app studyNot imported into the standard Library/app workflow
PRICE-02 rationale codingFinal 154-rationale transfer and coding pending
PRICE-02 statistical reportNumerical exact tests, price-model terms, mean intervals, decision-rule results and endpoint sensitivities available; code families and final signed record pending
This pagePublic review of the partial numerical results in the existing OpenPsy format
Numerical provenance

Counts, means, sample SDs, distributions and numerical inference are derived from the independently verified 153-row projection. The scalar tests use the unchanged source-pinned analysis functions; the separate numerical result seal is sha256:a926f9489071713b43eafe086008a1e5c580c481d3f116afd3546fa7ae377a13. No missing value is converted to zero. Exact source identities:

collectionSha: sha256:20becb2c1d023cee79efc0d4bdfb4a3a015661ae23db8de02303ef5161eb41db

descriptiveResultSha: sha256:5c818006af3f012b099b8218eb41ebe721839875dd59758a54bdc8c39e6cde86

independentAccountingSha: sha256:75d70b113ead3971d4824c329447b06355f566a479bc635b122afd7dbb5f8a4f

numericProjectionSha: sha256:5734f9611949b57777d180349c92327e6727de420e975572a89d4d4969adbdf0