It works ok, but I was wondering are the covariates centered around zero? I cannot find any specific function to do so in nilearn and the tutorial does not address this.
However, z-scoring is a simple way to center columns around zero (and also make all columns have unit variance, which is also nice for running models). That can be applied easily with
from scipy.stats import zscore
df.apply(zscore)
where df is a pandas dataframe that contains your design matrix.