Create matrix from a lagged timeseries numpy

Question

Say we have a time series

[1,2,3,4]

I would like to generate a matrix of lagged values (maybe using np.roll) that generates the matrix,

[