Summary Quick Dos And Donts

The field of biomarkers is a wide one, with a huge diversity of potential applications, all of which may have different and complex statistical analysis issues. This chapter has undoubtedly missed many of these and glossed over others, but summarizing concisely what it has covered is still a challenge. Instead, I present below a distillation in the form of one-liners addressing some of the more critical points ("do") and more common misconceptions ("don't").

1. Put on Hercule Poirot ' s hat and use your judgment when considering analytical issues and your experimental situation.

2. Think ahead: Consider data analysis issues before settling on your experimental design (and long before having the data in hand).

3. Take the limitations of your techniques and experimental error into account in your interpretations.

4. Check that your design is powered appropriately to detect what you wish to detect.

5. Use appropriate tests (i.e., ANOVA when comparing means from more than two groups).

6. Use appropriate controls and check HWE in genetic association studies.

7. Make adequate adjustment for the elevated false - positive rates when dealing with - omics - style data.

8. Average over multiple plausible candidate models with appropriate weights (rather than choosing a single one), for best predictive accuracy and uncertainty estimations in pharmacogenomic applications of Bayesian modeling strategies.

1. Use parametric tests if data do not meet the assumptions of these tests (such as being normally distributed).

2. Use repeated t-tests when comparing means from more than two groups in one experimental design.

3. Compare more groups than are relevant to the goals of your experiment when applying multiple means testing.

4. Use correlation analysis to infer cause and effect or to test agreement between different methods.

5. Simply equate the odds ratio and risk ratio without considering the outcome frequency.

6. Use data to test correlation to outcomes or as a test set for validation of algorithms if they were already involved with those earlier in the process of mathematical modeling of genomics data (i.e., selected based on outcomes in the former case, or involved in algorithm generation in the latter).

Project Management Made Easy

Project Management Made Easy

What you need to know about… Project Management Made Easy! Project management consists of more than just a large building project and can encompass small projects as well. No matter what the size of your project, you need to have some sort of project management. How you manage your project has everything to do with its outcome.

Get My Free Ebook

Post a comment