Because of sample variation, each SRL is at best an approximation of the true PRL.
We would get k different SRLs for k different samples and all these SRLs are not likely to be the same.
Then, we develop the concept of the sample regression function (SRF) to represent the SRL. It is written as