Research protocols
Proposed analyses of existing datasets. Each plan records a question, a method, a decision rule and the conditions for accessing the data.
Mental health consultations among Korean adults
How does past-year consultation with a professional differ between Korean adults meeting interview criteria for major depression and those meeting criteria for generalized anxiety disorder?
Method
Use the 2021 Korean adult survey to define mutually exclusive groups meeting past-year major depression criteria only, generalized anxiety criteria only, or both, using the K-CIDI flags MDD1Y and GAD1Y. Use past-year consultation with a professional (SR1C_1) as the outcome. Check past-year OCD and PTSD co-occurrence using OCD1Y and PTSD1Y. Apply survey strata, clusters and weights according to the codebook and user guide.
Decision criteria
The primary contrast is the percentage-point difference between the first two groups. Report a design-based 95% confidence interval and the difference after age and sex standardization. Suppress estimates for co-occurring groups that lack precision or fail the archive's disclosure requirements. Match the denominator and provider categories before checking the original paper's broader depression and anxiety totals.
Data and access
Limitations
The National Center for Mental Health must approve microdata access. The consultation item asks about seeking professional advice for a mental health issue. This comparison cannot establish disorder-specific care, treatment completion or the effect of consultation on symptoms.
Parent and teacher ratings in the MTA ADHD trial
How much does the 14-month difference in inattention between assignment to combined treatment and usual community care vary between parent and teacher ratings in the MTA trial?
Method
Compare the combined-treatment and community-care arms of the original MTA randomized sample. Use parent and teacher SNAP-IV inattention scores on the same scale at baseline and month 14. Analyze by original assignment, accounting for site and repeated observations within each child. Check baseline scores and month-14 missingness by arm and rater.
Decision criteria
Estimate the baseline-adjusted assignment contrast for each rater and the difference between the contrasts, with 95% confidence intervals. The difference is the primary estimand. Check whether conclusions change if missing month-14 scores are systematically higher or lower than observed scores.
Data and access
Limitations
NDA approval is required. The public catalogue lists SNAP-IV, while linkage of assignment, rater and visit must be confirmed in the approved package. Parents and teachers observe behavior in different settings. The contrast is specific to the original 14-month assignment period.
Head motion and functional connectivity in autism imaging
Does a resting-state connectivity contrast between autistic and comparison groups survive stricter head-motion exclusions and point in the same direction at independent sites?
Method
Use ABIDE I cross-sectional scans from participants aged 8–18. Analyze one resting-state correlation between medial prefrontal and posterior cingulate signals. Lock region definitions, preprocessing and motion calculation, then Fisher-transform the correlation. Estimate age- and sex-adjusted contrasts at sites containing both groups. Validate on ABIDE II cross-sectional data from institutions absent from ABIDE I and exclude follow-up scans of ABIDE I participants.
Decision criteria
Use mean framewise displacement of 0.20 mm or less as the primary threshold and 0.30 mm or less as a sensitivity analysis. Report group-specific exclusion rates, site-level contrasts and pooled 95% confidence intervals at both thresholds. Count independent-site evidence as supporting replication if the pooled estimate has the same direction and its interval excludes zero.
Data and access
Limitations
Confirm NITRC registration and data use terms. Acquisition timing and scanner differences can also affect motion metrics. Unequal exclusions can change the population represented by the estimate. This group contrast does not validate individual diagnosis.
Cognitive performance in families affected by psychosis
How does cognitive performance differ between independent controls and first-degree relatives without psychosis of people with schizophrenia or bipolar disorder with psychosis?
Method
In B-SNIP 1, distinguish index participants in the two diagnostic groups, their first-degree relatives without psychosis, and unrelated controls. Use the Brief Assessment of Cognition in Schizophrenia (BACS) composite. Account for age, sex, site and within-family dependence. First confirm that the approved package can identify the index diagnosis, link relatives and support scoring.
Decision criteria
Report adjusted standardized mean differences and 95% confidence intervals for each relative group versus unrelated controls. Estimate the difference between relative groups directly. Check sensitivity to selecting one relative per family using a rule fixed in advance.
Data and access
Limitations
NDA approval is required. Family-level analysis depends on access to family linkage information. Differences cannot separate shared environment from genetic influences, and they do not extend to bipolar disorder without psychosis.
Emotional support, sensory processing sensitivity and distress
Is the association between sensory processing sensitivity and psychological distress weaker at higher levels of emotional support?
Method
Use the unwinsorized release of U.S. adult data from the Multi-Sensitivity Study. Verify the HSP-12 mean, PROMIS Emotional Support Short Form 4a sum and Kessler Psychological Distress Scale (K10) sum against the codebook. Keep sensitivity and support continuous. Add one interaction to a model containing age, sex and both main effects.
Decision criteria
Estimate the difference in sensitivity–distress slopes at one standard deviation above versus below mean emotional support, subtracting the low-support slope from the high-support slope. The hypothesis is supported if the 95% confidence interval lies below zero and the direction is the same in the winsorized release.
Data and access
Limitations
Data and the codebook are open on OSF. Online self-reports collected at one time point cannot establish temporal order or a causal effect of support. K10 measures general psychological distress and HSP-12 measures a trait.