DOMStat can provide assistance with the logistics of study design and analysis.
Types of Study Designs
Systematic Literature Review/Meta-Analysis
Basic Science/lab study
Sample Size and Power
Determination of the appropriate sample size prior to study initiation is a crucial part of study design. The appropriate sample size for the planned study can be calculated using information from pilot studies or related studies in the same field. The study sample size should be large enough to be likely to detect important effects, but not so large that the study is longer and more costly than necessary.
Clinical trials generally require that patients/subjects to be randomized to treatments within sites in a double-blind manner. DOMStat can help design a randomization scheme, an approach to blinding, and a randomization plan and list. Randomization is also important for assigning lab animals to cages and treatments to avoid confounding of litter and cage placement effects with treatment results.
Planned Analysis Methods
The appropriate analysis methods depend on the study design and the type of outcome measures. Click on the study designs above for analysis methods often used for each type of study design. Below we have described analysis methods appropriate for various types of outcome measures and goals.
A meta-analysis is a literature review in which one tries to combine results across published studies. A detailed data collection form needs to be created at the beginning. This should include ratings of study quality (e.g. randomization and blinding methods). The biggest challenge, aside from locating relevant articles, is that different studies typically have different study designs, different outcome measures, use different time scales, and report different summary statistics. It is advisable to consult a statistician after the first reading of the selected articles, since all will probably need to be re-abstracted more than once.
Analysis methods focus on estimation of effect sizes, confidence intervals for effect sizes, and meta-regression modeling to assess the impact of fixed and/or random explanatory variables.