Biostatistics Resources

GITHUB

Visit the UNC CFAR Biostatistics GitHub at https://github.com/unc-cfar-bios.

Tools

R package for Serial Limiting Dilution Assay Statistics (SLDAssay)

Power Calculator for Repeated Low-Dose Challenge Studies

R Package for Rank-Sum Power Calculation (wmwpow)

Brief Statistical Safety Checklist

Purpose:

This checklist is designed to help review draft manuscripts, statistical reports, or analysis plans

Methods:

• Objective / hypotheses of the study are clearly stated
• Endpoints are defined, declare endpoints as primary, secondary, and exploratory as appropriate
• Exposures and covariates are defined
• Sample size calculation is explained (if applicable)
• If applicable, randomization and blinding process are explained
• If applicable, interim stopping guidelines are described
• All statistical methods are described (with references as needed)
• Methodological assumptions were reasonably met
• Reasonable considerations were made for:
  • missing data or attrition (e.g. loss to follow-up)
  • multiple testing or multiple comparisons
  • correlation among observations (e.g. longitudinal or clustered data)
  • small sample sizes (e.g. n<30)
• Statistical software and version used are stated (e.g. SAS version 9.3, Cary, North Carolina)
• Statistical type I error probability is declared (e.g. two-sided alpha of 0.05)
• Clinical trials: trial registration ID included (www.clinicaltrials.gov)

Results (including text, tables, figures):

• Number of decimal places reported appropriately reflects measurement precision
• Summaries of variability/spread are provided (e.g. SD, IQR)
• Sample sizes are provided (n=) and missing data are declared when applicable
• Each p-value is accompanied by a graphical display of data and descriptive statistics. Estimation with precision quantified (e.g., 95% confidence interval) is preferred.
• If reporting p-values, report specific p-values (p=0.027), rather than p<0.05
• Use of the word significant should be avoided
• Figures are clearly explained by a legend and all axes and symbols are appropriately labeled
• Distorted effect sizes (e.g. by truncation of y-axis) are clearly labeled and justified
• Units of measurement are stated

Discussion:

• Discussion conclusions are in agreement with results
• Limitations of the study design or analytic approach are appropriately stated

For more guidance, refer to standard reporting guidelines (e.g. CONSORT, STROBE, PRISMA, STARD) and the ASA p-value statement (and published updates thereof)

This list was compiled with reference to ‘Instructions for Authors’ at  JAMA, Annals of Internal Medicine, Nature, and Science Magazine.