In practice, the majority of the really big matrices I encounter are very sparse, because they're often constructed as a reduction from some other much smaller

In practice, the majority of the really big matrices I encounter are very sparse, because they're often constructed as a reduction from some other much smaller representation so we can use linear algebra methods.