zoom会议，需注册，美东时间。详情点阅读全文~

The objective of this workshop is to develop best practice recommendations for interpreting meaningful effects in “big data” by engaging scientists from a range of disciplines in discussions of meaningful science that go beyond statistical significance.

Wednesday, September 2, 2020 10:00 am – 6:00 pm ET

10:00 – 10:30 AM

Welcome & Opening Remarks

Elizabeth A. Hoffman, Ph.D., Scientific Program Manager, Adolescent Brain Cognitive DevelopmentSM Study, National Institute on Drug Abuse

Nora D. Volkow, M.D., Director, National Institute on Drug Abuse

George Koob, Ph.D., Director, National Institute on Alcohol Abuse and Alcoholism

Josh Gordon, M.D., Ph.D., Director, National Institute of Mental Health

10:30 – 11:00 AM

Keynote address, Jessica Wapner, Science Journalist, "Putting the reader first: How to communicate meaning in a meaningful way."

11:00 AM – 12:15 PM

Concept presentations

11:00 – 11:20 AM

Mike Hawrylycz, Ph.D., Allen Institute for Brain Science, “Meaningful effects in single-cell transcriptomics and epigenetics data”

11:20 – 11:40 AM

Ragini Verma, Ph.D., University of Pennsylvania, “Modeling meaningful effects in neuroimaging studies”

11:40 AM – 12:00 PM

Erin Dunn, Sc.D., Massachusetts General Hospital, “Identifying meaningful effects at the intersection between genes, life experiences, and development”

12:00 PM – 12:15 PM

Discussion

12:15 – 12:25 PM

Break

12:25 – 1:15 PM

Panel presentations

12:30 – 12:45 PM

Dana Hancock, Ph.D., RTI International, Small Effect Sizes. “Accumulating evidence from small effect sizes: examples in moving from genome-wide association studies to biology and clinical prediction”

12:45 – 1:00 PM

Vince Calhoun, Ph.D., Georgia Institute of Technology, Covariates/Collinearity. “Strategies to model data in the presence of confounds: examples from brain imaging”

1:00 – 1:15 PM

Jenn Pfeifer, Ph.D., University of Oregon, Exploratory, Confirmatory Frameworks. “The power of boundaries: confirmatory versus exploratory research in developmental and clinical neuroscience”

1:15 – 1:45 PM

Lunch

1:45 – 1:55 PM

Breakout Sessions Charge & Logistics

2:00 – 2:30 PM

Breakout Sessions I – Topics will be repeated for Breakout Sessions ll and lll. Participants will rotate through all three topics.

a. Small Effects:

• What is a “meaningful” effect? Even though differences may be highly statistically significant, the results may only account for a small proportion of the variance and/or have little ability to predict outcomes.

• How can small effect sizes be interpreted in terms of causality or prediction? For example, does a small effect size in an observational study necessarily mean that a subsequent experimental manipulation or intervention will not be effective, or could not serve as an accurate outcome predictor?

• Should there be different standards when interpreting results in terms of a detectable effect vs. an effect that could be the basis of an intervention?

• Effects may sit on the edge of a nonlinear inflection point so that a little movement in one variable causes disproportionate movement in another. When is a non-linear analysis justified in evaluating a small linear effect?

b. Covariates/Collinearity:

• Some variables have been traditionally viewed as confounds or nuisance variables; however, with large datasets, they may be more aptly incorporated into analytic models as variables of interest.

• Removing variance associated with one variable may impact other variables if the constructs are related.

• What is the role of the control variable in the underlying theoretical model? How does the exclusion/inclusion of certain control variables inform the model?

c. Exploratory vs. Confirmatory Data Analysis Frameworks:

• Distinguishing between the value of exploratory (e.g., effect size estimation) vs. confirmatory (hypothesis-driven) analytic approaches is especially important for emerging areas of study

• Exploratory approaches can inform confirmatory analyses, e.g., by building a strong base of effect size estimates to inform the development of a theoretical construct.

• Researcher degrees of freedom in confirmatory analyses (resulting from an extensive number of analysis decisions) can threaten inferences and impact Type 1 error.

• Pre-specification of analysis strategies via hypothesis pre-registration or registered reports enhances transparency and reproducibility.

2:35 – 3:05 PM

Breakout Sessions II

3:10 – 3:40 PM

Breakout Sessions III

3:45 – 4:10 PM

Break – breakout session facilitators prepare for report out.

4:10 – 4:45 PM

Report out from Breakouts

4:45 – 5:45 PM

Grand Discussion

5:45 – 6:00 PM

Wrap-up