Source: docs/R/01-figure-timing-assoc.R
The matlm R package fits linear models in matrix form and avoids calling lm.
That makes computation efficient if many predictors need to be tested,
while calling lm for every predictor results in a considerable overhead in computation time.
(yet to be implemented)
- Split a set of predictors into batches
- Parallel computation
- Support different formats to store predictors:
matrix,big.matrix, etc- Use S3 classes for the interface
- Run GLS models (not only OLS)