Deliberative distortions: available-wave sensitivity
Findings
No verdict policy supplied. Computed diagnostics do not imply PASS. Missing evidence does not establish absence of a problem.
- 129 event–item opinion changes are estimable: 71 positive, 57 negative and 1 zero. Signs follow each item's declared orientation; none is a quality verdict.
- Knowledge changes are estimable for 21 event–item combinations, ranging from 0.012 to 0.288 of the score range. These describe respondents with usable scores.
- 210 event–analysis combinations lack required inputs. The evidence table lists what would unlock each analysis.
Survey changes describe observed pre/post differences. They do not establish why views changed. Dialogue statistics describe recorded speech. Annotation results depend on supplied coding. D, P and H describe attitudes; their signs are not grades for deliberation.
Learning and opinion change
Points show separate event–item changes. Items need not measure the same construct; the report does not average them into a common success score. The main analysis uses the membership setting recorded below. Cluster or survey intervals appear in the diagnostic table only when their assumptions were explicitly declared.
H is the decrease in within-group attitude dispersion. Positive P is movement away from the baseline side’s midpoint. Positive D is movement toward the explicitly designated advantaged reference’s baseline view. Baseline ties can make a direction undefined.
Response coverage
Attrition is post-wave missingness among observed baseline respondents. It does not count people absent from both survey waves. Missingness can alter the population described by a change estimate; neither paired analysis nor robust standard errors establishes ignorable attrition.
Membership sensitivity
When an available-wave sensitivity run is supplied, this table compares it with the main event–item results. Differences can reflect changing respondent composition. Neither agreement nor disagreement establishes ignorable attrition.
No separate membership sensitivity run was supplied.
Participation and exchange
Speaking concentration and subgroup differences are diagnostics to investigate. Perfectly equal airtime is not imposed as a quality target. Argument coverage refers to the supplied inventory; it cannot establish that the inventory contains every important argument.
## NULL## NULLEvidence available
“Unavailable” means required inputs are missing. “Unassessable” means inputs exist but usable observations, definitions or inferential assumptions are insufficient. Each analysis lists its requirements.
Diagnostic explorer
Filter by event, item, metric or subgroup. CSV and Copy export the filtered rows. The table includes denominators, coverage, units, methods and reasons for unassessable results. Raw and Holm-adjusted p-values are separate; absent a declared testing family, tests are exploratory.