It is certainly a classical question but I haven't done stats for a long time. Let's say I have to do a Wald test (for a linear model) between the variable $Y$ and the variable $X_i$ with $i \in \{1,\dots,n\}.$ Since I'm doing $n$ tests, I use a threshold of $0.05/n$ for the $p$-value. However, these $X_i$ are not necessarily independent. Hence, I would like to estimate the "true" number of independent tests, let's say $n_0$ so that I could divise $0.05$ by $n_0$ instead of $n$. I guess I need to use the variance-covariance matrix of the $X_i$ to compute that $n_0$ but I cannot figure how exactly.
Thanks for any help.