-
Notifications
You must be signed in to change notification settings - Fork 317
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
controlling population stratification #2
Comments
Yes, they are correcting for systemic bias independent of the trait in the sense that any association or lack thereof has no bearing on the MDS calculations. Once could substitute Principal Component Analysis PCA at this stage for MDS, as I understand it, the only difference is in the projection to 2D. If they aren't associated with the trait, then they shouldn't change the outcome much anyway but should be included as a matter of principle. Still, statistical significance is probably too high of a bar to set here and there are likely effects that are noteworthy despite not being significant. Perhaps a better test of whether or not to add them is the relative effect of each subsequent component, i.e. if you consider 10 MDS components vs. 20. At some point there is usually a large fall off (at least if it is like PCA, which I'm pretty sure it is) and it isn't really useful to include more components beyond that point. |
Hi,
Thank you for the paper and the step-by-step tutorial! Those are of great help!
May I ask a question regarding your paper? It says that the following:
Does it mean those MDS components should be added as covariates regardless whether they are significantly associated with the trait? if so, could you explain why?
The text was updated successfully, but these errors were encountered: