Ellimination of insignificant parameters in multipe linear regression with reglinr function

[samuel.enibe at unn.edu.ng]

I have a set of data X made up of n columns. I wish to use the reglinr linear regression function to correlate the data with Y, such that Y =f(X).

What is the best  function or procedure to 
minimise the number of columns of X needed to descibe the function while maximising the R2 coeffient.

Samuel Enibe