Speeding up subsetting of data.table and implementation of thousands of regression

Question

I have a data.table with 100 rows and 3 columns. The rows are grouped into 30 groups. The three columns are my independent variables.

During each iteration,