How does the Lasso function handle NaN values?

I have a dataset (assembled as a matrix) that I want to run perform LASSO regression on, however, as a processing step outliers are removed and marked as NaN.

My question is how does the LASSO function handle these NaN values, would it be worth me performing some level of imputation?

Answer 1

[David Fink]

The lasso function ignores all rows (observations) with NaN values.

This can be seen in the code for 'lasso' near line 240 where it checks isfinite() on each row.

>> open lasso